Parent Directory
|
Revision Log
Revision 982 - (view) (download)
1 : | edgomez | 200 | /***************************************************************************** |
2 : | edgomez | 201 | * |
3 : | * XVID MPEG-4 VIDEO CODEC | ||
4 : | * - Native API implementation - | ||
5 : | * | ||
6 : | edgomez | 851 | * This program is free software ; you can redistribute it and/or modify |
7 : | * it under the terms of the GNU General Public License as published by | ||
8 : | * the Free Software Foundation ; either version 2 of the License, or | ||
9 : | edgomez | 201 | * (at your option) any later version. |
10 : | * | ||
11 : | * This program is distributed in the hope that it will be useful, | ||
12 : | edgomez | 851 | * but WITHOUT ANY WARRANTY ; without even the implied warranty of |
13 : | edgomez | 201 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
14 : | * GNU General Public License for more details. | ||
15 : | * | ||
16 : | * You should have received a copy of the GNU General Public License | ||
17 : | edgomez | 851 | * along with this program ; if not, write to the Free Software |
18 : | edgomez | 201 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
19 : | * | ||
20 : | edgomez | 982 | * $Id: xvid.c,v 1.45.2.3 2003-04-10 13:05:54 edgomez Exp $ |
21 : | edgomez | 410 | * |
22 : | edgomez | 201 | ****************************************************************************/ |
23 : | chenm001 | 274 | |
24 : | edgomez | 851 | #include <stdio.h> |
25 : | #include <stdlib.h> | ||
26 : | #include <string.h> | ||
27 : | #include <time.h> | ||
28 : | |||
29 : | Isibaar | 3 | #include "xvid.h" |
30 : | #include "decoder.h" | ||
31 : | #include "encoder.h" | ||
32 : | #include "bitstream/cbp.h" | ||
33 : | #include "dct/idct.h" | ||
34 : | #include "dct/fdct.h" | ||
35 : | #include "image/colorspace.h" | ||
36 : | #include "image/interpolate8x8.h" | ||
37 : | edgomez | 851 | #include "image/reduced.h" |
38 : | Isibaar | 3 | #include "utils/mem_transfer.h" |
39 : | edgomez | 851 | #include "utils/mbfunctions.h" |
40 : | Isibaar | 3 | #include "quant/quant_h263.h" |
41 : | #include "quant/quant_mpeg4.h" | ||
42 : | ia64p | 299 | #include "motion/motion.h" |
43 : | Isibaar | 3 | #include "motion/sad.h" |
44 : | #include "utils/emms.h" | ||
45 : | #include "utils/timer.h" | ||
46 : | Isibaar | 100 | #include "bitstream/mbcoding.h" |
47 : | Isibaar | 3 | |
48 : | edgomez | 851 | #if defined(ARCH_IS_IA32) |
49 : | suxen_drol | 311 | |
50 : | edgomez | 851 | #if defined(_MSC_VER) |
51 : | # include <windows.h> | ||
52 : | suxen_drol | 311 | #else |
53 : | edgomez | 851 | # include <signal.h> |
54 : | # include <setjmp.h> | ||
55 : | suxen_drol | 311 | |
56 : | edgomez | 851 | static jmp_buf mark; |
57 : | suxen_drol | 311 | |
58 : | edgomez | 851 | static void |
59 : | sigill_handler(int signal) | ||
60 : | { | ||
61 : | longjmp(mark, 1); | ||
62 : | } | ||
63 : | suxen_drol | 311 | #endif |
64 : | |||
65 : | |||
66 : | /* | ||
67 : | edgomez | 874 | * Calls the funcptr, and returns whether SIGILL (illegal instruction) was |
68 : | * signalled | ||
69 : | * | ||
70 : | * Return values: | ||
71 : | * -1 : could not determine | ||
72 : | * 0 : SIGILL was *not* signalled | ||
73 : | * 1 : SIGILL was signalled | ||
74 : | */ | ||
75 : | suxen_drol | 311 | |
76 : | int | ||
77 : | sigill_check(void (*func)()) | ||
78 : | { | ||
79 : | edgomez | 851 | #if defined(_MSC_VER) |
80 : | suxen_drol | 311 | _try { |
81 : | func(); | ||
82 : | } | ||
83 : | _except(EXCEPTION_EXECUTE_HANDLER) { | ||
84 : | |||
85 : | if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION) | ||
86 : | return 1; | ||
87 : | } | ||
88 : | return 0; | ||
89 : | #else | ||
90 : | void * old_handler; | ||
91 : | int jmpret; | ||
92 : | |||
93 : | |||
94 : | old_handler = signal(SIGILL, sigill_handler); | ||
95 : | if (old_handler == SIG_ERR) | ||
96 : | { | ||
97 : | return -1; | ||
98 : | } | ||
99 : | |||
100 : | jmpret = setjmp(mark); | ||
101 : | if (jmpret == 0) | ||
102 : | { | ||
103 : | func(); | ||
104 : | } | ||
105 : | |||
106 : | signal(SIGILL, old_handler); | ||
107 : | |||
108 : | return jmpret; | ||
109 : | #endif | ||
110 : | } | ||
111 : | #endif | ||
112 : | |||
113 : | edgomez | 851 | |
114 : | /* detect cpu flags */ | ||
115 : | static unsigned int | ||
116 : | detect_cpu_flags() | ||
117 : | { | ||
118 : | /* enable native assembly optimizations by default */ | ||
119 : | unsigned int cpu_flags = XVID_CPU_ASM; | ||
120 : | |||
121 : | #if defined(ARCH_IS_IA32) | ||
122 : | cpu_flags |= check_cpu_features(); | ||
123 : | if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger)) | ||
124 : | cpu_flags &= ~XVID_CPU_SSE; | ||
125 : | |||
126 : | if ((cpu_flags & XVID_CPU_SSE2) && sigill_check(sse2_os_trigger)) | ||
127 : | cpu_flags &= ~XVID_CPU_SSE2; | ||
128 : | #endif | ||
129 : | |||
130 : | #if defined(ARCH_IS_PPC) | ||
131 : | #if defined(ARCH_IS_PPC_ALTIVEC) | ||
132 : | cpu_flags |= XVID_CPU_ALTIVEC; | ||
133 : | #endif | ||
134 : | #endif | ||
135 : | |||
136 : | return cpu_flags; | ||
137 : | } | ||
138 : | |||
139 : | |||
140 : | edgomez | 200 | /***************************************************************************** |
141 : | * XviD Init Entry point | ||
142 : | * | ||
143 : | * Well this function initialize all internal function pointers according | ||
144 : | * to the CPU features forced by the library client or autodetected (depending | ||
145 : | * on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all | ||
146 : | * image colorspace transformation tables. | ||
147 : | * | ||
148 : | * Returned value : XVID_ERR_OK | ||
149 : | * + API_VERSION in the input XVID_INIT_PARAM structure | ||
150 : | * + core build " " " " " | ||
151 : | * | ||
152 : | ****************************************************************************/ | ||
153 : | |||
154 : | edgomez | 851 | |
155 : | static | ||
156 : | suxen_drol | 890 | int xvid_gbl_init(xvid_gbl_init_t * init) |
157 : | Isibaar | 3 | { |
158 : | suxen_drol | 890 | unsigned int cpu_flags; |
159 : | Isibaar | 3 | |
160 : | suxen_drol | 890 | if (XVID_MAJOR(init->version) != 1) /* v1.x.x */ |
161 : | return XVID_ERR_VERSION; | ||
162 : | suxen_drol | 234 | |
163 : | suxen_drol | 890 | cpu_flags = (init->cpu_flags & XVID_CPU_FORCE) ? init->cpu_flags : detect_cpu_flags(); |
164 : | suxen_drol | 234 | |
165 : | edgomez | 200 | /* Initialize the function pointers */ |
166 : | Isibaar | 3 | idct_int32_init(); |
167 : | Isibaar | 100 | init_vlc_tables(); |
168 : | |||
169 : | edgomez | 200 | /* Fixed Point Forward/Inverse DCT transformations */ |
170 : | Isibaar | 3 | fdct = fdct_int32; |
171 : | idct = idct_int32; | ||
172 : | |||
173 : | edgomez | 200 | /* Only needed on PPC Altivec archs */ |
174 : | canard | 115 | sadInit = 0; |
175 : | edgomez | 195 | |
176 : | edgomez | 200 | /* Restore FPU context : emms_c is a nop functions */ |
177 : | Isibaar | 3 | emms = emms_c; |
178 : | |||
179 : | edgomez | 200 | /* Quantization functions */ |
180 : | quant_intra = quant_intra_c; | ||
181 : | Isibaar | 3 | dequant_intra = dequant_intra_c; |
182 : | edgomez | 200 | quant_inter = quant_inter_c; |
183 : | Isibaar | 3 | dequant_inter = dequant_inter_c; |
184 : | |||
185 : | edgomez | 200 | quant4_intra = quant4_intra_c; |
186 : | Isibaar | 3 | dequant4_intra = dequant4_intra_c; |
187 : | edgomez | 200 | quant4_inter = quant4_inter_c; |
188 : | Isibaar | 3 | dequant4_inter = dequant4_inter_c; |
189 : | |||
190 : | edgomez | 200 | /* Block transfer related functions */ |
191 : | Isibaar | 3 | transfer_8to16copy = transfer_8to16copy_c; |
192 : | transfer_16to8copy = transfer_16to8copy_c; | ||
193 : | edgomez | 200 | transfer_8to16sub = transfer_8to16sub_c; |
194 : | edgomez | 851 | transfer_8to16subro = transfer_8to16subro_c; |
195 : | suxen_drol | 118 | transfer_8to16sub2 = transfer_8to16sub2_c; |
196 : | edgomez | 200 | transfer_16to8add = transfer_16to8add_c; |
197 : | transfer8x8_copy = transfer8x8_copy_c; | ||
198 : | Isibaar | 3 | |
199 : | edgomez | 851 | /* Interlacing functions */ |
200 : | MBFieldTest = MBFieldTest_c; | ||
201 : | |||
202 : | edgomez | 200 | /* Image interpolation related functions */ |
203 : | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_c; | ||
204 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; | ||
205 : | Isibaar | 3 | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
206 : | |||
207 : | edgomez | 851 | interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c; |
208 : | interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c; | ||
209 : | interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c; | ||
210 : | |||
211 : | interpolate8x8_lowpass_h = interpolate8x8_lowpass_h_c; | ||
212 : | interpolate8x8_lowpass_v = interpolate8x8_lowpass_v_c; | ||
213 : | interpolate8x8_lowpass_hv = interpolate8x8_lowpass_hv_c; | ||
214 : | |||
215 : | interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_c; | ||
216 : | interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_c; | ||
217 : | |||
218 : | interpolate8x8_avg2 = interpolate8x8_avg2_c; | ||
219 : | interpolate8x8_avg4 = interpolate8x8_avg4_c; | ||
220 : | |||
221 : | /* reduced resoltuion */ | ||
222 : | copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_C; | ||
223 : | add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_C; | ||
224 : | vfilter_31 = xvid_VFilter_31_C; | ||
225 : | hfilter_31 = xvid_HFilter_31_C; | ||
226 : | filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_C; | ||
227 : | filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_C; | ||
228 : | |||
229 : | edgomez | 200 | /* Initialize internal colorspace transformation tables */ |
230 : | Isibaar | 3 | colorspace_init(); |
231 : | |||
232 : | edgomez | 200 | /* All colorspace transformation functions User Format->YV12 */ |
233 : | edgomez | 851 | yv12_to_yv12 = yv12_to_yv12_c; |
234 : | rgb555_to_yv12 = rgb555_to_yv12_c; | ||
235 : | rgb565_to_yv12 = rgb565_to_yv12_c; | ||
236 : | bgr_to_yv12 = bgr_to_yv12_c; | ||
237 : | bgra_to_yv12 = bgra_to_yv12_c; | ||
238 : | abgr_to_yv12 = abgr_to_yv12_c; | ||
239 : | rgba_to_yv12 = rgba_to_yv12_c; | ||
240 : | yuyv_to_yv12 = yuyv_to_yv12_c; | ||
241 : | uyvy_to_yv12 = uyvy_to_yv12_c; | ||
242 : | Isibaar | 3 | |
243 : | edgomez | 851 | rgb555i_to_yv12 = rgb555i_to_yv12_c; |
244 : | rgb565i_to_yv12 = rgb565i_to_yv12_c; | ||
245 : | bgri_to_yv12 = bgri_to_yv12_c; | ||
246 : | bgrai_to_yv12 = bgrai_to_yv12_c; | ||
247 : | abgri_to_yv12 = abgri_to_yv12_c; | ||
248 : | rgbai_to_yv12 = rgbai_to_yv12_c; | ||
249 : | yuyvi_to_yv12 = yuyvi_to_yv12_c; | ||
250 : | uyvyi_to_yv12 = uyvyi_to_yv12_c; | ||
251 : | |||
252 : | |||
253 : | edgomez | 200 | /* All colorspace transformation functions YV12->User format */ |
254 : | edgomez | 851 | yv12_to_rgb555 = yv12_to_rgb555_c; |
255 : | yv12_to_rgb565 = yv12_to_rgb565_c; | ||
256 : | yv12_to_bgr = yv12_to_bgr_c; | ||
257 : | yv12_to_bgra = yv12_to_bgra_c; | ||
258 : | yv12_to_abgr = yv12_to_abgr_c; | ||
259 : | yv12_to_rgba = yv12_to_rgba_c; | ||
260 : | yv12_to_yuyv = yv12_to_yuyv_c; | ||
261 : | yv12_to_uyvy = yv12_to_uyvy_c; | ||
262 : | |||
263 : | yv12_to_rgb555i = yv12_to_rgb555i_c; | ||
264 : | yv12_to_rgb565i = yv12_to_rgb565i_c; | ||
265 : | yv12_to_bgri = yv12_to_bgri_c; | ||
266 : | yv12_to_bgrai = yv12_to_bgrai_c; | ||
267 : | yv12_to_abgri = yv12_to_abgri_c; | ||
268 : | yv12_to_rgbai = yv12_to_rgbai_c; | ||
269 : | yv12_to_yuyvi = yv12_to_yuyvi_c; | ||
270 : | yv12_to_uyvyi = yv12_to_uyvyi_c; | ||
271 : | Isibaar | 3 | |
272 : | edgomez | 200 | /* Functions used in motion estimation algorithms */ |
273 : | Isibaar | 3 | calc_cbp = calc_cbp_c; |
274 : | edgomez | 200 | sad16 = sad16_c; |
275 : | suxen_drol | 329 | sad8 = sad8_c; |
276 : | edgomez | 200 | sad16bi = sad16bi_c; |
277 : | suxen_drol | 329 | sad8bi = sad8bi_c; |
278 : | edgomez | 200 | dev16 = dev16_c; |
279 : | edgomez | 851 | sad16v = sad16v_c; |
280 : | suxen_drol | 329 | |
281 : | edgomez | 874 | /* Halfpel8_Refine = Halfpel8_Refine_c; */ |
282 : | Isibaar | 3 | |
283 : | edgomez | 851 | #if defined(ARCH_IS_IA32) |
284 : | edgomez | 200 | |
285 : | edgomez | 851 | if ((cpu_flags & XVID_CPU_ASM)) |
286 : | { | ||
287 : | vfilter_31 = xvid_VFilter_31_x86; | ||
288 : | hfilter_31 = xvid_HFilter_31_x86; | ||
289 : | } | ||
290 : | |||
291 : | if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || | ||
292 : | (cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || | ||
293 : | (cpu_flags & XVID_CPU_SSE) || (cpu_flags & XVID_CPU_SSE2)) | ||
294 : | { | ||
295 : | /* Restore FPU context : emms_c is a nop functions */ | ||
296 : | emms = emms_mmx; | ||
297 : | } | ||
298 : | |||
299 : | if ((cpu_flags & XVID_CPU_MMX)) { | ||
300 : | |||
301 : | edgomez | 200 | /* Forward and Inverse Discrete Cosine Transformation functions */ |
302 : | Isibaar | 3 | fdct = fdct_mmx; |
303 : | idct = idct_mmx; | ||
304 : | |||
305 : | edgomez | 200 | /* Quantization related functions */ |
306 : | quant_intra = quant_intra_mmx; | ||
307 : | Isibaar | 3 | dequant_intra = dequant_intra_mmx; |
308 : | edgomez | 200 | quant_inter = quant_inter_mmx; |
309 : | Isibaar | 3 | dequant_inter = dequant_inter_mmx; |
310 : | |||
311 : | edgomez | 200 | quant4_intra = quant4_intra_mmx; |
312 : | Isibaar | 3 | dequant4_intra = dequant4_intra_mmx; |
313 : | edgomez | 200 | quant4_inter = quant4_inter_mmx; |
314 : | Isibaar | 3 | dequant4_inter = dequant4_inter_mmx; |
315 : | |||
316 : | edgomez | 200 | /* Block related functions */ |
317 : | Isibaar | 3 | transfer_8to16copy = transfer_8to16copy_mmx; |
318 : | transfer_16to8copy = transfer_16to8copy_mmx; | ||
319 : | edgomez | 200 | transfer_8to16sub = transfer_8to16sub_mmx; |
320 : | edgomez | 851 | transfer_8to16subro = transfer_8to16subro_mmx; |
321 : | edgomez | 236 | transfer_8to16sub2 = transfer_8to16sub2_mmx; |
322 : | edgomez | 200 | transfer_16to8add = transfer_16to8add_mmx; |
323 : | transfer8x8_copy = transfer8x8_copy_mmx; | ||
324 : | Isibaar | 3 | |
325 : | edgomez | 851 | /* Interlacing Functions */ |
326 : | MBFieldTest = MBFieldTest_mmx; | ||
327 : | edgomez | 236 | |
328 : | edgomez | 200 | /* Image Interpolation related functions */ |
329 : | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx; | ||
330 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; | ||
331 : | Isibaar | 3 | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
332 : | |||
333 : | edgomez | 851 | interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; |
334 : | interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; | ||
335 : | |||
336 : | interpolate8x8_avg2 = interpolate8x8_avg2_mmx; | ||
337 : | interpolate8x8_avg4 = interpolate8x8_avg4_mmx; | ||
338 : | |||
339 : | /* reduced resolution */ | ||
340 : | copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_mmx; | ||
341 : | add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_mmx; | ||
342 : | hfilter_31 = xvid_HFilter_31_mmx; | ||
343 : | filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_mmx; | ||
344 : | filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_mmx; | ||
345 : | |||
346 : | /* image input xxx_to_yv12 related functions */ | ||
347 : | yv12_to_yv12 = yv12_to_yv12_mmx; | ||
348 : | bgr_to_yv12 = bgr_to_yv12_mmx; | ||
349 : | bgra_to_yv12 = bgra_to_yv12_mmx; | ||
350 : | edgomez | 200 | yuyv_to_yv12 = yuyv_to_yv12_mmx; |
351 : | uyvy_to_yv12 = uyvy_to_yv12_mmx; | ||
352 : | Isibaar | 3 | |
353 : | edgomez | 851 | /* image output yv12_to_xxx related functions */ |
354 : | yv12_to_bgr = yv12_to_bgr_mmx; | ||
355 : | yv12_to_bgra = yv12_to_bgra_mmx; | ||
356 : | edgomez | 200 | yv12_to_yuyv = yv12_to_yuyv_mmx; |
357 : | yv12_to_uyvy = yv12_to_uyvy_mmx; | ||
358 : | edgomez | 851 | |
359 : | yv12_to_yuyvi = yv12_to_yuyvi_mmx; | ||
360 : | yv12_to_uyvyi = yv12_to_uyvyi_mmx; | ||
361 : | Isibaar | 3 | |
362 : | edgomez | 200 | /* Motion estimation related functions */ |
363 : | Isibaar | 3 | calc_cbp = calc_cbp_mmx; |
364 : | edgomez | 200 | sad16 = sad16_mmx; |
365 : | sad8 = sad8_mmx; | ||
366 : | suxen_drol | 329 | sad16bi = sad16bi_mmx; |
367 : | sad8bi = sad8bi_mmx; | ||
368 : | edgomez | 200 | dev16 = dev16_mmx; |
369 : | edgomez | 851 | sad16v = sad16v_mmx; |
370 : | Isibaar | 3 | } |
371 : | |||
372 : | suxen_drol | 329 | /* these 3dnow functions are faster than mmx, but slower than xmm. */ |
373 : | edgomez | 851 | if ((cpu_flags & XVID_CPU_3DNOW)) { |
374 : | suxen_drol | 329 | |
375 : | edgomez | 851 | emms = emms_3dn; |
376 : | |||
377 : | suxen_drol | 329 | /* ME functions */ |
378 : | sad16bi = sad16bi_3dn; | ||
379 : | sad8bi = sad8bi_3dn; | ||
380 : | edgomez | 851 | |
381 : | yuyv_to_yv12 = yuyv_to_yv12_3dn; | ||
382 : | uyvy_to_yv12 = uyvy_to_yv12_3dn; | ||
383 : | suxen_drol | 329 | } |
384 : | |||
385 : | |||
386 : | edgomez | 851 | if ((cpu_flags & XVID_CPU_MMXEXT)) { |
387 : | edgomez | 200 | |
388 : | /* Inverse DCT */ | ||
389 : | Isibaar | 3 | idct = idct_xmm; |
390 : | edgomez | 200 | |
391 : | /* Interpolation */ | ||
392 : | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_xmm; | ||
393 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; | ||
394 : | h | 38 | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
395 : | edgomez | 200 | |
396 : | edgomez | 851 | /* reduced resolution */ |
397 : | copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_xmm; | ||
398 : | add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_xmm; | ||
399 : | |||
400 : | chenm001 | 274 | /* Quantization */ |
401 : | edgomez | 851 | quant4_intra = quant4_intra_xmm; |
402 : | quant4_inter = quant4_inter_xmm; | ||
403 : | |||
404 : | chenm001 | 274 | dequant_intra = dequant_intra_xmm; |
405 : | dequant_inter = dequant_inter_xmm; | ||
406 : | |||
407 : | edgomez | 218 | /* Buffer transfer */ |
408 : | transfer_8to16sub2 = transfer_8to16sub2_xmm; | ||
409 : | |||
410 : | edgomez | 200 | /* Colorspace transformation */ |
411 : | edgomez | 851 | yv12_to_yv12 = yv12_to_yv12_xmm; |
412 : | yuyv_to_yv12 = yuyv_to_yv12_xmm; | ||
413 : | uyvy_to_yv12 = uyvy_to_yv12_xmm; | ||
414 : | Isibaar | 3 | |
415 : | edgomez | 200 | /* ME functions */ |
416 : | Isibaar | 3 | sad16 = sad16_xmm; |
417 : | suxen_drol | 329 | sad8 = sad8_xmm; |
418 : | chenm001 | 274 | sad16bi = sad16bi_xmm; |
419 : | suxen_drol | 329 | sad8bi = sad8bi_xmm; |
420 : | Isibaar | 3 | dev16 = dev16_xmm; |
421 : | edgomez | 851 | sad16v = sad16v_xmm; |
422 : | Isibaar | 3 | } |
423 : | |||
424 : | edgomez | 851 | if ((cpu_flags & XVID_CPU_3DNOW)) { |
425 : | edgomez | 200 | |
426 : | /* Interpolation */ | ||
427 : | Isibaar | 3 | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn; |
428 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dn; | ||
429 : | h | 40 | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn; |
430 : | Isibaar | 3 | } |
431 : | |||
432 : | edgomez | 851 | if ((cpu_flags & XVID_CPU_3DNOWEXT)) { |
433 : | edgomez | 200 | |
434 : | edgomez | 851 | /* Inverse DCT */ |
435 : | idct = idct_3dne; | ||
436 : | |||
437 : | /* Buffer transfer */ | ||
438 : | transfer_8to16copy = transfer_8to16copy_3dne; | ||
439 : | transfer_16to8copy = transfer_16to8copy_3dne; | ||
440 : | transfer_8to16sub = transfer_8to16sub_3dne; | ||
441 : | transfer_8to16subro = transfer_8to16subro_3dne; | ||
442 : | transfer_8to16sub2 = transfer_8to16sub2_3dne; | ||
443 : | transfer_16to8add = transfer_16to8add_3dne; | ||
444 : | transfer8x8_copy = transfer8x8_copy_3dne; | ||
445 : | |||
446 : | /* Quantization */ | ||
447 : | dequant4_intra = dequant4_intra_3dne; | ||
448 : | dequant4_inter = dequant4_inter_3dne; | ||
449 : | quant_intra = quant_intra_3dne; | ||
450 : | quant_inter = quant_inter_3dne; | ||
451 : | dequant_intra = dequant_intra_3dne; | ||
452 : | dequant_inter = dequant_inter_3dne; | ||
453 : | |||
454 : | /* ME functions */ | ||
455 : | calc_cbp = calc_cbp_3dne; | ||
456 : | sad16 = sad16_3dne; | ||
457 : | sad8 = sad8_3dne; | ||
458 : | sad16bi = sad16bi_3dne; | ||
459 : | sad8bi = sad8bi_3dne; | ||
460 : | dev16 = dev16_3dne; | ||
461 : | |||
462 : | /* Interpolation */ | ||
463 : | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne; | ||
464 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne; | ||
465 : | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne; | ||
466 : | } | ||
467 : | |||
468 : | |||
469 : | if ((cpu_flags & XVID_CPU_SSE2)) { | ||
470 : | |||
471 : | chenm001 | 274 | calc_cbp = calc_cbp_sse2; |
472 : | |||
473 : | edgomez | 200 | /* Quantization */ |
474 : | quant_intra = quant_intra_sse2; | ||
475 : | Isibaar | 154 | dequant_intra = dequant_intra_sse2; |
476 : | edgomez | 200 | quant_inter = quant_inter_sse2; |
477 : | Isibaar | 154 | dequant_inter = dequant_inter_sse2; |
478 : | h | 135 | |
479 : | edgomez | 851 | #if defined(EXPERIMENTAL_SSE2_CODE) |
480 : | /* ME; slower than xmm */ | ||
481 : | edgomez | 200 | sad16 = sad16_sse2; |
482 : | dev16 = dev16_sse2; | ||
483 : | edgomez | 851 | #endif |
484 : | edgomez | 200 | /* Forward and Inverse DCT */ |
485 : | idct = idct_sse2; | ||
486 : | Isibaar | 154 | fdct = fdct_sse2; |
487 : | h | 126 | } |
488 : | Isibaar | 3 | #endif |
489 : | edgomez | 200 | |
490 : | edgomez | 851 | #if defined(ARCH_IS_IA64) |
491 : | edgomez | 874 | if ((cpu_flags & XVID_CPU_ASM)) { /* use assembler routines? */ |
492 : | Isibaar | 209 | idct_ia64_init(); |
493 : | fdct = fdct_ia64; | ||
494 : | edgomez | 874 | idct = idct_ia64; /*not yet working, crashes */ |
495 : | Isibaar | 209 | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64; |
496 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64; | ||
497 : | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64; | ||
498 : | sad16 = sad16_ia64; | ||
499 : | sad16bi = sad16bi_ia64; | ||
500 : | sad8 = sad8_ia64; | ||
501 : | dev16 = dev16_ia64; | ||
502 : | edgomez | 874 | /* Halfpel8_Refine = Halfpel8_Refine_ia64; */ |
503 : | Isibaar | 209 | quant_intra = quant_intra_ia64; |
504 : | dequant_intra = dequant_intra_ia64; | ||
505 : | quant_inter = quant_inter_ia64; | ||
506 : | dequant_inter = dequant_inter_ia64; | ||
507 : | transfer_8to16copy = transfer_8to16copy_ia64; | ||
508 : | transfer_16to8copy = transfer_16to8copy_ia64; | ||
509 : | transfer_8to16sub = transfer_8to16sub_ia64; | ||
510 : | transfer_8to16sub2 = transfer_8to16sub2_ia64; | ||
511 : | transfer_16to8add = transfer_16to8add_ia64; | ||
512 : | transfer8x8_copy = transfer8x8_copy_ia64; | ||
513 : | edgomez | 874 | DPRINTF(DPRINTF_DEBUG, "Using IA-64 assembler routines."); |
514 : | Isibaar | 209 | } |
515 : | #endif | ||
516 : | |||
517 : | edgomez | 851 | #if defined(ARCH_IS_PPC) |
518 : | if ((cpu_flags & XVID_CPU_ASM)) | ||
519 : | { | ||
520 : | calc_cbp = calc_cbp_ppc; | ||
521 : | } | ||
522 : | |||
523 : | if ((cpu_flags & XVID_CPU_ALTIVEC)) | ||
524 : | { | ||
525 : | calc_cbp = calc_cbp_altivec; | ||
526 : | fdct = fdct_altivec; | ||
527 : | idct = idct_altivec; | ||
528 : | sadInit = sadInit_altivec; | ||
529 : | sad16 = sad16_altivec; | ||
530 : | sad8 = sad8_altivec; | ||
531 : | dev16 = dev16_altivec; | ||
532 : | } | ||
533 : | canard | 52 | #endif |
534 : | edgomez | 851 | |
535 : | suxen_drol | 890 | return 0; |
536 : | edgomez | 851 | } |
537 : | |||
538 : | |||
539 : | suxen_drol | 890 | static int |
540 : | xvid_gbl_info(xvid_gbl_info_t * info) | ||
541 : | { | ||
542 : | if (XVID_MAJOR(info->version) != 1) /* v1.x.x */ | ||
543 : | return XVID_ERR_VERSION; | ||
544 : | edgomez | 851 | |
545 : | suxen_drol | 890 | info->actual_version = XVID_VERSION; |
546 : | info->build = "dev-api-4"; | ||
547 : | info->cpu_flags = detect_cpu_flags(); | ||
548 : | |||
549 : | #if defined(_SMP) && defined(WIN32) | ||
550 : | info->num_threads = pthread_num_processors_np();; | ||
551 : | #else | ||
552 : | info->num_threads = 0; | ||
553 : | #endif | ||
554 : | |||
555 : | return 0; | ||
556 : | } | ||
557 : | |||
558 : | |||
559 : | edgomez | 851 | static int |
560 : | suxen_drol | 890 | xvid_gbl_convert(xvid_gbl_convert_t* convert) |
561 : | edgomez | 851 | { |
562 : | suxen_drol | 890 | int width; |
563 : | int height; | ||
564 : | int width2; | ||
565 : | int height2; | ||
566 : | edgomez | 851 | IMAGE img; |
567 : | |||
568 : | suxen_drol | 890 | if (XVID_MAJOR(convert->version) != 1) /* v1.x.x */ |
569 : | return XVID_ERR_VERSION; | ||
570 : | |||
571 : | // const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP); | ||
572 : | width = convert->width; | ||
573 : | height = convert->height; | ||
574 : | width2 = convert->width/2; | ||
575 : | height2 = convert->height/2; | ||
576 : | |||
577 : | switch (convert->input.csp & ~XVID_CSP_VFLIP) | ||
578 : | edgomez | 851 | { |
579 : | case XVID_CSP_YV12 : | ||
580 : | suxen_drol | 890 | img.y = convert->input.plane[0]; |
581 : | img.v = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height; | ||
582 : | img.u = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height + (convert->input.stride[0]/2)*height2; | ||
583 : | edgomez | 851 | image_output(&img, width, height, width, |
584 : | suxen_drol | 890 | (uint8_t**)convert->output.plane, convert->output.stride, |
585 : | convert->output.csp, convert->interlacing); | ||
586 : | edgomez | 851 | break; |
587 : | |||
588 : | default : | ||
589 : | return XVID_ERR_FORMAT; | ||
590 : | } | ||
591 : | |||
592 : | |||
593 : | emms(); | ||
594 : | suxen_drol | 890 | return 0; |
595 : | edgomez | 851 | } |
596 : | |||
597 : | |||
598 : | |||
599 : | void fill8(uint8_t * block, int size, int value) | ||
600 : | { | ||
601 : | int i; | ||
602 : | for (i = 0; i < size; i++) | ||
603 : | block[i] = value; | ||
604 : | } | ||
605 : | |||
606 : | void fill16(int16_t * block, int size, int value) | ||
607 : | { | ||
608 : | int i; | ||
609 : | for (i = 0; i < size; i++) | ||
610 : | block[i] = value; | ||
611 : | } | ||
612 : | |||
613 : | #define RANDOM(min,max) min + (rand() % (max-min)) | ||
614 : | |||
615 : | void random8(uint8_t * block, int size, int min, int max) | ||
616 : | { | ||
617 : | int i; | ||
618 : | for (i = 0; i < size; i++) | ||
619 : | block[i] = RANDOM(min,max); | ||
620 : | } | ||
621 : | |||
622 : | void random16(int16_t * block, int size, int min, int max) | ||
623 : | { | ||
624 : | int i; | ||
625 : | for (i = 0; i < size; i++) | ||
626 : | block[i] = RANDOM(min,max); | ||
627 : | } | ||
628 : | |||
629 : | int compare16(const int16_t * blockA, const int16_t * blockB, int size) | ||
630 : | { | ||
631 : | int i; | ||
632 : | for (i = 0; i < size; i++) | ||
633 : | if (blockA[i] != blockB[i]) | ||
634 : | return 1; | ||
635 : | |||
636 : | return 0; | ||
637 : | } | ||
638 : | |||
639 : | int diff16(const int16_t * blockA, const int16_t * blockB, int size) | ||
640 : | { | ||
641 : | int i, diff = 0; | ||
642 : | for (i = 0; i < size; i++) | ||
643 : | edgomez | 982 | diff += abs(blockA[i]-blockB[i]); |
644 : | edgomez | 851 | return diff; |
645 : | } | ||
646 : | |||
647 : | |||
648 : | #define XVID_TEST_RANDOM 0x00000001 /* random input data */ | ||
649 : | #define XVID_TEST_VERBOSE 0x00000002 /* verbose error output */ | ||
650 : | |||
651 : | |||
652 : | #define TEST_FORWARD 0x00000001 /* intra */ | ||
653 : | #define TEST_FDCT (TEST_FORWARD) | ||
654 : | #define TEST_IDCT (0) | ||
655 : | |||
656 : | suxen_drol | 860 | static int test_transform(void * funcA, void * funcB, const char * nameB, |
657 : | edgomez | 851 | int test, int flags) |
658 : | { | ||
659 : | int i; | ||
660 : | int64_t timeSTART; | ||
661 : | int64_t timeA = 0; | ||
662 : | int64_t timeB = 0; | ||
663 : | DECLARE_ALIGNED_MATRIX(arrayA, 1, 64, int16_t, CACHE_LINE); | ||
664 : | DECLARE_ALIGNED_MATRIX(arrayB, 1, 64, int16_t, CACHE_LINE); | ||
665 : | int min, max; | ||
666 : | int count = 0; | ||
667 : | |||
668 : | int tmp; | ||
669 : | int min_error = 0x10000*64; | ||
670 : | int max_error = 0; | ||
671 : | |||
672 : | |||
673 : | if ((test & TEST_FORWARD)) /* forward */ | ||
674 : | { | ||
675 : | min = -256; | ||
676 : | max = 255; | ||
677 : | }else{ /* inverse */ | ||
678 : | min = -2048; | ||
679 : | max = 2047; | ||
680 : | } | ||
681 : | |||
682 : | for (i = 0; i < 64*64; i++) | ||
683 : | { | ||
684 : | if ((flags & XVID_TEST_RANDOM)) | ||
685 : | { | ||
686 : | random16(arrayA, 64, min, max); | ||
687 : | }else{ | ||
688 : | fill16(arrayA, 64, i); | ||
689 : | } | ||
690 : | memcpy(arrayB, arrayA, 64*sizeof(int16_t)); | ||
691 : | |||
692 : | if ((test & TEST_FORWARD)) | ||
693 : | { | ||
694 : | timeSTART = read_counter(); | ||
695 : | ((fdctFunc*)funcA)(arrayA); | ||
696 : | timeA += read_counter() - timeSTART; | ||
697 : | |||
698 : | timeSTART = read_counter(); | ||
699 : | ((fdctFunc*)funcB)(arrayB); | ||
700 : | timeB += read_counter() - timeSTART; | ||
701 : | } | ||
702 : | else | ||
703 : | { | ||
704 : | timeSTART = read_counter(); | ||
705 : | ((idctFunc*)funcA)(arrayA); | ||
706 : | timeA += read_counter() - timeSTART; | ||
707 : | |||
708 : | timeSTART = read_counter(); | ||
709 : | ((idctFunc*)funcB)(arrayB); | ||
710 : | timeB += read_counter() - timeSTART; | ||
711 : | } | ||
712 : | |||
713 : | tmp = diff16(arrayA, arrayB, 64) / 64; | ||
714 : | if (tmp > max_error) | ||
715 : | max_error = tmp; | ||
716 : | if (tmp < min_error) | ||
717 : | min_error = tmp; | ||
718 : | |||
719 : | count++; | ||
720 : | } | ||
721 : | |||
722 : | /* print the "average difference" of best/worst transforms */ | ||
723 : | printf("%s:\t%i\t(min_error:%i, max_error:%i)\n", nameB, (int)(timeB / count), min_error, max_error); | ||
724 : | |||
725 : | return 0; | ||
726 : | } | ||
727 : | |||
728 : | |||
729 : | #define TEST_QUANT 0x00000001 /* forward quantization */ | ||
730 : | #define TEST_INTRA 0x00000002 /* intra */ | ||
731 : | #define TEST_QUANT_INTRA (TEST_QUANT|TEST_INTRA) | ||
732 : | #define TEST_QUANT_INTER (TEST_QUANT) | ||
733 : | #define TEST_DEQUANT_INTRA (TEST_INTRA) | ||
734 : | #define TEST_DEQUANT_INTER (0) | ||
735 : | |||
736 : | suxen_drol | 860 | static int test_quant(void * funcA, void * funcB, const char * nameB, |
737 : | edgomez | 851 | int test, int flags) |
738 : | { | ||
739 : | int q,i; | ||
740 : | int64_t timeSTART; | ||
741 : | int64_t timeA = 0; | ||
742 : | int64_t timeB = 0; | ||
743 : | edgomez | 874 | int retA = 0, retB = 0; |
744 : | edgomez | 851 | DECLARE_ALIGNED_MATRIX(arrayX, 1, 64, int16_t, CACHE_LINE); |
745 : | DECLARE_ALIGNED_MATRIX(arrayA, 1, 64, int16_t, CACHE_LINE); | ||
746 : | DECLARE_ALIGNED_MATRIX(arrayB, 1, 64, int16_t, CACHE_LINE); | ||
747 : | int min, max; | ||
748 : | int count = 0; | ||
749 : | int errors = 0; | ||
750 : | |||
751 : | if ((test & TEST_QUANT)) /* quant */ | ||
752 : | { | ||
753 : | min = -2048; | ||
754 : | max = 2047; | ||
755 : | }else{ /* dequant */ | ||
756 : | min = -256; | ||
757 : | max = 255; | ||
758 : | } | ||
759 : | |||
760 : | for (q = 1; q <= 31; q++) /* quantizer */ | ||
761 : | { | ||
762 : | for (i = min; i < max; i++) /* input coeff */ | ||
763 : | { | ||
764 : | if ((flags & XVID_TEST_RANDOM)) | ||
765 : | { | ||
766 : | random16(arrayX, 64, min, max); | ||
767 : | }else{ | ||
768 : | fill16(arrayX, 64, i); | ||
769 : | } | ||
770 : | |||
771 : | if ((test & TEST_INTRA)) /* intra */ | ||
772 : | { | ||
773 : | timeSTART = read_counter(); | ||
774 : | ((quanth263_intraFunc*)funcA)(arrayA, arrayX, q, q); | ||
775 : | timeA += read_counter() - timeSTART; | ||
776 : | |||
777 : | timeSTART = read_counter(); | ||
778 : | ((quanth263_intraFunc*)funcB)(arrayB, arrayX, q, q); | ||
779 : | timeB += read_counter() - timeSTART; | ||
780 : | } | ||
781 : | else /* inter */ | ||
782 : | { | ||
783 : | timeSTART = read_counter(); | ||
784 : | retA = ((quanth263_interFunc*)funcA)(arrayA, arrayX, q); | ||
785 : | timeA += read_counter() - timeSTART; | ||
786 : | |||
787 : | timeSTART = read_counter(); | ||
788 : | retB = ((quanth263_interFunc*)funcB)(arrayB, arrayX, q); | ||
789 : | timeB += read_counter() - timeSTART; | ||
790 : | } | ||
791 : | |||
792 : | /* compare return value from quant_inter, and compare (de)quantiz'd arrays */ | ||
793 : | if ( ((test&TEST_QUANT) && !(test&TEST_INTRA) && retA != retB ) || | ||
794 : | compare16(arrayA, arrayB, 64)) | ||
795 : | { | ||
796 : | errors++; | ||
797 : | if ((flags & XVID_TEST_VERBOSE)) | ||
798 : | printf("%s error: q=%i, i=%i\n", nameB, q, i); | ||
799 : | } | ||
800 : | |||
801 : | count++; | ||
802 : | } | ||
803 : | } | ||
804 : | |||
805 : | printf("%s:\t%i", nameB, (int)(timeB / count)); | ||
806 : | if (errors>0) | ||
807 : | printf("\t(%i errors out of %i)", errors, count); | ||
808 : | printf("\n"); | ||
809 : | |||
810 : | return 0; | ||
811 : | } | ||
812 : | |||
813 : | |||
814 : | |||
815 : | int xvid_init_test(int flags) | ||
816 : | { | ||
817 : | edgomez | 882 | #if defined(ARCH_IS_IA32) |
818 : | int cpu_flags; | ||
819 : | #endif | ||
820 : | edgomez | 851 | |
821 : | edgomez | 879 | printf("XviD tests\n\n"); |
822 : | edgomez | 851 | |
823 : | edgomez | 879 | #if defined(ARCH_IS_IA32) |
824 : | edgomez | 851 | cpu_flags = detect_cpu_flags(); |
825 : | edgomez | 879 | #endif |
826 : | |||
827 : | edgomez | 851 | idct_int32_init(); |
828 : | edgomez | 879 | emms(); |
829 : | edgomez | 851 | |
830 : | edgomez | 882 | srand(time(0)); |
831 : | |||
832 : | edgomez | 879 | /* fDCT test */ |
833 : | edgomez | 851 | printf("--- fdct ---\n"); |
834 : | test_transform(fdct_int32, fdct_int32, "c", TEST_FDCT, flags); | ||
835 : | edgomez | 879 | |
836 : | #if defined(ARCH_IS_IA32) | ||
837 : | edgomez | 851 | if (cpu_flags & XVID_CPU_MMX) |
838 : | test_transform(fdct_int32, fdct_mmx, "mmx", TEST_FDCT, flags); | ||
839 : | if (cpu_flags & XVID_CPU_SSE2) | ||
840 : | test_transform(fdct_int32, fdct_sse2, "sse2", TEST_FDCT, flags); | ||
841 : | edgomez | 879 | #endif |
842 : | edgomez | 851 | |
843 : | edgomez | 879 | /* iDCT test */ |
844 : | edgomez | 851 | printf("\n--- idct ---\n"); |
845 : | test_transform(idct_int32, idct_int32, "c", TEST_IDCT, flags); | ||
846 : | edgomez | 879 | |
847 : | #if defined(ARCH_IS_IA32) | ||
848 : | edgomez | 851 | if (cpu_flags & XVID_CPU_MMX) |
849 : | test_transform(idct_int32, idct_mmx, "mmx", TEST_IDCT, flags); | ||
850 : | if (cpu_flags & XVID_CPU_MMXEXT) | ||
851 : | test_transform(idct_int32, idct_xmm, "xmm", TEST_IDCT, flags); | ||
852 : | if (cpu_flags & XVID_CPU_3DNOWEXT) | ||
853 : | test_transform(idct_int32, idct_3dne, "3dne", TEST_IDCT, flags); | ||
854 : | if (cpu_flags & XVID_CPU_SSE2) | ||
855 : | test_transform(idct_int32, idct_sse2, "sse2", TEST_IDCT, flags); | ||
856 : | edgomez | 879 | #endif |
857 : | edgomez | 851 | |
858 : | edgomez | 879 | /* Intra quantization test */ |
859 : | edgomez | 851 | printf("\n--- quant intra ---\n"); |
860 : | test_quant(quant_intra_c, quant_intra_c, "c", TEST_QUANT_INTRA, flags); | ||
861 : | edgomez | 879 | |
862 : | #if defined(ARCH_IS_IA32) | ||
863 : | edgomez | 851 | if (cpu_flags & XVID_CPU_MMX) |
864 : | test_quant(quant_intra_c, quant_intra_mmx, "mmx", TEST_QUANT_INTRA, flags); | ||
865 : | if (cpu_flags & XVID_CPU_3DNOWEXT) | ||
866 : | test_quant(quant_intra_c, quant_intra_3dne, "3dne", TEST_QUANT_INTRA, flags); | ||
867 : | if (cpu_flags & XVID_CPU_SSE2) | ||
868 : | test_quant(quant_intra_c, quant_intra_sse2, "sse2", TEST_QUANT_INTRA, flags); | ||
869 : | edgomez | 879 | #endif |
870 : | edgomez | 851 | |
871 : | edgomez | 879 | /* Inter quantization test */ |
872 : | edgomez | 851 | printf("\n--- quant inter ---\n"); |
873 : | test_quant(quant_inter_c, quant_inter_c, "c", TEST_QUANT_INTER, flags); | ||
874 : | edgomez | 879 | |
875 : | #if defined(ARCH_IS_IA32) | ||
876 : | edgomez | 851 | if (cpu_flags & XVID_CPU_MMX) |
877 : | test_quant(quant_inter_c, quant_inter_mmx, "mmx", TEST_QUANT_INTER, flags); | ||
878 : | if (cpu_flags & XVID_CPU_3DNOWEXT) | ||
879 : | test_quant(quant_inter_c, quant_inter_3dne, "3dne", TEST_QUANT_INTER, flags); | ||
880 : | if (cpu_flags & XVID_CPU_SSE2) | ||
881 : | test_quant(quant_inter_c, quant_inter_sse2, "sse2", TEST_QUANT_INTER, flags); | ||
882 : | edgomez | 879 | #endif |
883 : | edgomez | 851 | |
884 : | edgomez | 879 | /* Intra dequantization test */ |
885 : | edgomez | 851 | printf("\n--- dequant intra ---\n"); |
886 : | test_quant(dequant_intra_c, dequant_intra_c, "c", TEST_DEQUANT_INTRA, flags); | ||
887 : | edgomez | 879 | |
888 : | #if defined(ARCH_IS_IA32) | ||
889 : | edgomez | 851 | if (cpu_flags & XVID_CPU_MMX) |
890 : | test_quant(dequant_intra_c, dequant_intra_mmx, "mmx", TEST_DEQUANT_INTRA, flags); | ||
891 : | if (cpu_flags & XVID_CPU_MMXEXT) | ||
892 : | test_quant(dequant_intra_c, dequant_intra_xmm, "xmm", TEST_DEQUANT_INTRA, flags); | ||
893 : | if (cpu_flags & XVID_CPU_3DNOWEXT) | ||
894 : | test_quant(dequant_intra_c, dequant_intra_3dne, "3dne", TEST_DEQUANT_INTRA, flags); | ||
895 : | if (cpu_flags & XVID_CPU_SSE2) | ||
896 : | test_quant(dequant_intra_c, dequant_intra_sse2, "sse2", TEST_DEQUANT_INTRA, flags); | ||
897 : | edgomez | 879 | #endif |
898 : | edgomez | 851 | |
899 : | edgomez | 879 | /* Inter dequantization test */ |
900 : | edgomez | 851 | printf("\n--- dequant inter ---\n"); |
901 : | test_quant(dequant_inter_c, dequant_inter_c, "c", TEST_DEQUANT_INTER, flags); | ||
902 : | edgomez | 879 | |
903 : | #if defined(ARCH_IS_IA32) | ||
904 : | edgomez | 851 | if (cpu_flags & XVID_CPU_MMX) |
905 : | test_quant(dequant_inter_c, dequant_inter_mmx, "mmx", TEST_DEQUANT_INTER, flags); | ||
906 : | if (cpu_flags & XVID_CPU_MMXEXT) | ||
907 : | test_quant(dequant_inter_c, dequant_inter_xmm, "xmm", TEST_DEQUANT_INTER, flags); | ||
908 : | if (cpu_flags & XVID_CPU_3DNOWEXT) | ||
909 : | test_quant(dequant_inter_c, dequant_inter_3dne, "3dne", TEST_DEQUANT_INTER, flags); | ||
910 : | if (cpu_flags & XVID_CPU_SSE2) | ||
911 : | test_quant(dequant_inter_c, dequant_inter_sse2, "sse2", TEST_DEQUANT_INTER, flags); | ||
912 : | edgomez | 879 | #endif |
913 : | edgomez | 851 | |
914 : | edgomez | 879 | /* Intra quantization test */ |
915 : | printf("\n--- quant4 intra ---\n"); | ||
916 : | edgomez | 851 | test_quant(quant4_intra_c, quant4_intra_c, "c", TEST_QUANT_INTRA, flags); |
917 : | edgomez | 879 | |
918 : | #if defined(ARCH_IS_IA32) | ||
919 : | edgomez | 851 | if (cpu_flags & XVID_CPU_MMX) |
920 : | test_quant(quant4_intra_c, quant4_intra_mmx, "mmx", TEST_QUANT_INTRA, flags); | ||
921 : | if (cpu_flags & XVID_CPU_MMXEXT) | ||
922 : | test_quant(quant4_intra_c, quant4_intra_xmm, "xmm", TEST_QUANT_INTRA, flags); | ||
923 : | edgomez | 879 | #endif |
924 : | edgomez | 851 | |
925 : | edgomez | 879 | /* Inter quantization test */ |
926 : | printf("\n--- quant4 inter ---\n"); | ||
927 : | edgomez | 851 | test_quant(quant4_inter_c, quant4_inter_c, "c", TEST_QUANT_INTER, flags); |
928 : | edgomez | 879 | |
929 : | #if defined(ARCH_IS_IA32) | ||
930 : | edgomez | 851 | if (cpu_flags & XVID_CPU_MMX) |
931 : | test_quant(quant4_inter_c, quant4_inter_mmx, "mmx", TEST_QUANT_INTER, flags); | ||
932 : | if (cpu_flags & XVID_CPU_MMXEXT) | ||
933 : | test_quant(quant4_inter_c, quant4_inter_xmm, "xmm", TEST_QUANT_INTER, flags); | ||
934 : | edgomez | 879 | #endif |
935 : | edgomez | 851 | |
936 : | edgomez | 879 | /* Intra dequantization test */ |
937 : | printf("\n--- dequant4 intra ---\n"); | ||
938 : | edgomez | 851 | test_quant(dequant4_intra_c, dequant4_intra_c, "c", TEST_DEQUANT_INTRA, flags); |
939 : | edgomez | 879 | |
940 : | #if defined(ARCH_IS_IA32) | ||
941 : | edgomez | 851 | if (cpu_flags & XVID_CPU_MMX) |
942 : | test_quant(dequant4_intra_c, dequant4_intra_mmx, "mmx", TEST_DEQUANT_INTRA, flags); | ||
943 : | if (cpu_flags & XVID_CPU_3DNOWEXT) | ||
944 : | test_quant(dequant4_intra_c, dequant4_intra_3dne, "3dne", TEST_DEQUANT_INTRA, flags); | ||
945 : | edgomez | 879 | #endif |
946 : | edgomez | 851 | |
947 : | edgomez | 879 | /* Inter dequantization test */ |
948 : | printf("\n--- dequant4 inter ---\n"); | ||
949 : | edgomez | 851 | test_quant(dequant4_inter_c, dequant4_inter_c, "c", TEST_DEQUANT_INTER, flags); |
950 : | edgomez | 879 | |
951 : | #if defined(ARCH_IS_IA32) | ||
952 : | edgomez | 851 | if (cpu_flags & XVID_CPU_MMX) |
953 : | test_quant(dequant4_inter_c, dequant4_inter_mmx, "mmx", TEST_DEQUANT_INTER, flags); | ||
954 : | if (cpu_flags & XVID_CPU_3DNOWEXT) | ||
955 : | test_quant(dequant4_inter_c, dequant4_inter_3dne, "3dne", TEST_DEQUANT_INTER, flags); | ||
956 : | edgomez | 879 | #endif |
957 : | edgomez | 851 | |
958 : | edgomez | 879 | emms(); |
959 : | edgomez | 851 | |
960 : | suxen_drol | 890 | return 0; |
961 : | Isibaar | 3 | } |
962 : | |||
963 : | edgomez | 851 | |
964 : | suxen_drol | 890 | /***************************************************************************** |
965 : | * XviD Global Entry point | ||
966 : | * | ||
967 : | * Well this function initialize all internal function pointers according | ||
968 : | * to the CPU features forced by the library client or autodetected (depending | ||
969 : | * on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all | ||
970 : | * image colorspace transformation tables. | ||
971 : | * | ||
972 : | ****************************************************************************/ | ||
973 : | |||
974 : | |||
975 : | edgomez | 851 | int |
976 : | suxen_drol | 890 | xvid_global(void *handle, |
977 : | edgomez | 851 | int opt, |
978 : | void *param1, | ||
979 : | void *param2) | ||
980 : | { | ||
981 : | switch(opt) | ||
982 : | { | ||
983 : | suxen_drol | 890 | case XVID_GBL_INIT : |
984 : | return xvid_gbl_init((xvid_gbl_init_t*)param1); | ||
985 : | edgomez | 851 | |
986 : | suxen_drol | 890 | case XVID_GBL_INFO : |
987 : | return xvid_gbl_info((xvid_gbl_info_t*)param1); | ||
988 : | edgomez | 851 | |
989 : | suxen_drol | 890 | case XVID_GBL_CONVERT : |
990 : | return xvid_gbl_convert((xvid_gbl_convert_t*)param1); | ||
991 : | |||
992 : | case XVID_GBL_TEST : | ||
993 : | edgomez | 879 | { |
994 : | ptr_t flags = (ptr_t)param1; | ||
995 : | return xvid_init_test((int)flags); | ||
996 : | } | ||
997 : | edgomez | 851 | default : |
998 : | return XVID_ERR_FAIL; | ||
999 : | } | ||
1000 : | } | ||
1001 : | |||
1002 : | edgomez | 200 | /***************************************************************************** |
1003 : | * XviD Native decoder entry point | ||
1004 : | * | ||
1005 : | * This function is just a wrapper to all the option cases. | ||
1006 : | * | ||
1007 : | * Returned values : XVID_ERR_FAIL when opt is invalid | ||
1008 : | * else returns the wrapped function result | ||
1009 : | * | ||
1010 : | ****************************************************************************/ | ||
1011 : | |||
1012 : | edgomez | 195 | int |
1013 : | xvid_decore(void *handle, | ||
1014 : | int opt, | ||
1015 : | void *param1, | ||
1016 : | void *param2) | ||
1017 : | Isibaar | 3 | { |
1018 : | edgomez | 195 | switch (opt) { |
1019 : | case XVID_DEC_CREATE: | ||
1020 : | suxen_drol | 890 | return decoder_create((xvid_dec_create_t *) param1); |
1021 : | Isibaar | 3 | |
1022 : | edgomez | 195 | case XVID_DEC_DESTROY: |
1023 : | return decoder_destroy((DECODER *) handle); | ||
1024 : | |||
1025 : | suxen_drol | 890 | case XVID_DEC_DECODE: |
1026 : | return decoder_decode((DECODER *) handle, (xvid_dec_frame_t *) param1, (xvid_dec_stats_t*) param2); | ||
1027 : | |||
1028 : | Isibaar | 3 | default: |
1029 : | edgomez | 195 | return XVID_ERR_FAIL; |
1030 : | } | ||
1031 : | Isibaar | 3 | } |
1032 : | |||
1033 : | |||
1034 : | edgomez | 200 | /***************************************************************************** |
1035 : | * XviD Native encoder entry point | ||
1036 : | * | ||
1037 : | * This function is just a wrapper to all the option cases. | ||
1038 : | * | ||
1039 : | * Returned values : XVID_ERR_FAIL when opt is invalid | ||
1040 : | * else returns the wrapped function result | ||
1041 : | * | ||
1042 : | ****************************************************************************/ | ||
1043 : | |||
1044 : | edgomez | 195 | int |
1045 : | xvid_encore(void *handle, | ||
1046 : | int opt, | ||
1047 : | void *param1, | ||
1048 : | void *param2) | ||
1049 : | Isibaar | 3 | { |
1050 : | edgomez | 195 | switch (opt) { |
1051 : | case XVID_ENC_ENCODE: | ||
1052 : | edgomez | 851 | |
1053 : | suxen_drol | 890 | return enc_encode((Encoder *) handle, |
1054 : | (xvid_enc_frame_t *) param1, | ||
1055 : | (xvid_enc_stats_t *) param2); | ||
1056 : | Isibaar | 3 | |
1057 : | edgomez | 195 | case XVID_ENC_CREATE: |
1058 : | suxen_drol | 948 | return enc_create((xvid_enc_create_t *) param1); |
1059 : | Isibaar | 3 | |
1060 : | edgomez | 195 | case XVID_ENC_DESTROY: |
1061 : | suxen_drol | 890 | return enc_destroy((Encoder *) handle); |
1062 : | edgomez | 195 | |
1063 : | Isibaar | 3 | default: |
1064 : | edgomez | 195 | return XVID_ERR_FAIL; |
1065 : | } | ||
1066 : | Isibaar | 3 | } |
No admin address has been configured | ViewVC Help |
Powered by ViewVC 1.0.4 |