37 |
* - 22.12.2001 API change: added xvid_init() - Isibaar |
* - 22.12.2001 API change: added xvid_init() - Isibaar |
38 |
* - 16.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
* - 16.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
39 |
* |
* |
40 |
* $Id: xvid.c,v 1.27 2002-07-09 01:37:21 chenm001 Exp $ |
* $Id: xvid.c,v 1.33.2.7 2002-10-06 07:05:50 suxen_drol Exp $ |
41 |
* |
* |
42 |
****************************************************************************/ |
****************************************************************************/ |
43 |
|
|
50 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
51 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
52 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
53 |
|
#include "utils/mbfunctions.h" |
54 |
#include "quant/quant_h263.h" |
#include "quant/quant_h263.h" |
55 |
#include "quant/quant_mpeg4.h" |
#include "quant/quant_mpeg4.h" |
56 |
|
#include "motion/motion.h" |
57 |
#include "motion/sad.h" |
#include "motion/sad.h" |
58 |
#include "utils/emms.h" |
#include "utils/emms.h" |
59 |
#include "utils/timer.h" |
#include "utils/timer.h" |
60 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
61 |
|
|
62 |
|
#if defined(ARCH_X86) && defined(EXPERIMENTAL_SSE2_CODE) |
63 |
|
|
64 |
|
#ifdef WIN32 |
65 |
|
#include <windows.h> |
66 |
|
#else |
67 |
|
#include <signal.h> |
68 |
|
#include <setjmp.h> |
69 |
|
#endif |
70 |
|
|
71 |
|
|
72 |
|
#ifndef WIN32 |
73 |
|
|
74 |
|
static jmp_buf mark; |
75 |
|
|
76 |
|
static void |
77 |
|
sigill_handler(int signal) |
78 |
|
{ |
79 |
|
longjmp(mark, 1); |
80 |
|
} |
81 |
|
#endif |
82 |
|
|
83 |
|
|
84 |
|
/* |
85 |
|
calls the funcptr, and returns whether SIGILL (illegal instruction) was signalled |
86 |
|
return values: |
87 |
|
-1 : could not determine |
88 |
|
0 : SIGILL was *not* signalled |
89 |
|
1 : SIGILL was signalled |
90 |
|
*/ |
91 |
|
|
92 |
|
int |
93 |
|
sigill_check(void (*func)()) |
94 |
|
{ |
95 |
|
#ifdef WIN32 |
96 |
|
_try { |
97 |
|
func(); |
98 |
|
} |
99 |
|
_except(EXCEPTION_EXECUTE_HANDLER) { |
100 |
|
|
101 |
|
if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION) |
102 |
|
return 1; |
103 |
|
} |
104 |
|
return 0; |
105 |
|
#else |
106 |
|
void * old_handler; |
107 |
|
int jmpret; |
108 |
|
|
109 |
|
|
110 |
|
old_handler = signal(SIGILL, sigill_handler); |
111 |
|
if (old_handler == SIG_ERR) |
112 |
|
{ |
113 |
|
return -1; |
114 |
|
} |
115 |
|
|
116 |
|
jmpret = setjmp(mark); |
117 |
|
if (jmpret == 0) |
118 |
|
{ |
119 |
|
func(); |
120 |
|
} |
121 |
|
|
122 |
|
signal(SIGILL, old_handler); |
123 |
|
|
124 |
|
return jmpret; |
125 |
|
#endif |
126 |
|
} |
127 |
|
#endif |
128 |
|
|
129 |
/***************************************************************************** |
/***************************************************************************** |
130 |
* XviD Init Entry point |
* XviD Init Entry point |
131 |
* |
* |
157 |
/* Inform the client the core build - unused because we're still alpha */ |
/* Inform the client the core build - unused because we're still alpha */ |
158 |
init_param->core_build = 1000; |
init_param->core_build = 1000; |
159 |
|
|
|
if ((init_param->cpu_flags & XVID_CPU_CHKONLY)) |
|
|
{ |
|
|
init_param->cpu_flags = check_cpu_features(); |
|
|
return XVID_ERR_OK; |
|
|
} |
|
|
|
|
160 |
/* Do we have to force CPU features ? */ |
/* Do we have to force CPU features ? */ |
161 |
if ((init_param->cpu_flags & XVID_CPU_FORCE) > 0) { |
if ((init_param->cpu_flags & XVID_CPU_FORCE)) { |
162 |
|
|
163 |
cpu_flags = init_param->cpu_flags; |
cpu_flags = init_param->cpu_flags; |
164 |
|
|
165 |
} else { |
} else { |
166 |
|
|
167 |
cpu_flags = check_cpu_features(); |
cpu_flags = check_cpu_features(); |
168 |
|
|
169 |
|
#if defined(ARCH_X86) && defined(EXPERIMENTAL_SSE2_CODE) |
170 |
|
if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger)) |
171 |
|
cpu_flags &= ~XVID_CPU_SSE; |
172 |
|
|
173 |
|
if ((cpu_flags & XVID_CPU_SSE2) && sigill_check(sse2_os_trigger)) |
174 |
|
cpu_flags &= ~XVID_CPU_SSE2; |
175 |
|
#endif |
176 |
|
} |
177 |
|
|
178 |
|
if ((init_param->cpu_flags & XVID_CPU_CHKONLY)) |
179 |
|
{ |
180 |
init_param->cpu_flags = cpu_flags; |
init_param->cpu_flags = cpu_flags; |
181 |
|
return XVID_ERR_OK; |
182 |
} |
} |
183 |
|
|
184 |
|
init_param->cpu_flags = cpu_flags; |
185 |
|
|
186 |
|
|
187 |
/* Initialize the function pointers */ |
/* Initialize the function pointers */ |
188 |
idct_int32_init(); |
idct_int32_init(); |
189 |
init_vlc_tables(); |
init_vlc_tables(); |
217 |
transfer_16to8add = transfer_16to8add_c; |
transfer_16to8add = transfer_16to8add_c; |
218 |
transfer8x8_copy = transfer8x8_copy_c; |
transfer8x8_copy = transfer8x8_copy_c; |
219 |
|
|
220 |
|
/* Interlacing functions */ |
221 |
|
MBFieldTest = MBFieldTest_c; |
222 |
|
|
223 |
/* Image interpolation related functions */ |
/* Image interpolation related functions */ |
224 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_c; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_c; |
225 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
226 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
227 |
|
|
228 |
|
interpolate8x8_lowpass_h = interpolate8x8_lowpass_h_c; |
229 |
|
interpolate8x8_lowpass_v = interpolate8x8_lowpass_v_c; |
230 |
|
interpolate8x8_lowpass_hv = interpolate8x8_lowpass_hv_c; |
231 |
|
|
232 |
|
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_c; |
233 |
|
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_c; |
234 |
|
|
235 |
|
interpolate8x8_avg2 = interpolate8x8_avg2_c; |
236 |
|
interpolate8x8_avg4 = interpolate8x8_avg4_c; |
237 |
|
|
238 |
/* Initialize internal colorspace transformation tables */ |
/* Initialize internal colorspace transformation tables */ |
239 |
colorspace_init(); |
colorspace_init(); |
240 |
|
|
252 |
yv12_to_rgb565 = yv12_to_rgb565_c; |
yv12_to_rgb565 = yv12_to_rgb565_c; |
253 |
yv12_to_rgb24 = yv12_to_rgb24_c; |
yv12_to_rgb24 = yv12_to_rgb24_c; |
254 |
yv12_to_rgb32 = yv12_to_rgb32_c; |
yv12_to_rgb32 = yv12_to_rgb32_c; |
255 |
|
yv12_to_abgr = yv12_to_abgr_c; |
256 |
|
yv12_to_rgba = yv12_to_rgba_c; |
257 |
yv12_to_yuv = yv12_to_yuv_c; |
yv12_to_yuv = yv12_to_yuv_c; |
258 |
yv12_to_yuyv = yv12_to_yuyv_c; |
yv12_to_yuyv = yv12_to_yuyv_c; |
259 |
yv12_to_uyvy = yv12_to_uyvy_c; |
yv12_to_uyvy = yv12_to_uyvy_c; |
261 |
/* Functions used in motion estimation algorithms */ |
/* Functions used in motion estimation algorithms */ |
262 |
calc_cbp = calc_cbp_c; |
calc_cbp = calc_cbp_c; |
263 |
sad16 = sad16_c; |
sad16 = sad16_c; |
|
sad16bi = sad16bi_c; |
|
264 |
sad8 = sad8_c; |
sad8 = sad8_c; |
265 |
|
sad16bi = sad16bi_c; |
266 |
|
sad8bi = sad8bi_c; |
267 |
dev16 = dev16_c; |
dev16 = dev16_c; |
268 |
|
sad16v = sad16v_c; |
269 |
|
|
270 |
|
// Halfpel8_Refine = Halfpel8_Refine_c; |
271 |
|
|
272 |
#ifdef ARCH_X86 |
#ifdef ARCH_X86 |
273 |
if ((cpu_flags & XVID_CPU_MMX) > 0) { |
if ((cpu_flags & XVID_CPU_MMX) > 0) { |
298 |
transfer_16to8add = transfer_16to8add_mmx; |
transfer_16to8add = transfer_16to8add_mmx; |
299 |
transfer8x8_copy = transfer8x8_copy_mmx; |
transfer8x8_copy = transfer8x8_copy_mmx; |
300 |
|
|
301 |
|
/* Interlacing Functions */ |
302 |
|
MBFieldTest = MBFieldTest_mmx; |
303 |
|
|
304 |
/* Image Interpolation related functions */ |
/* Image Interpolation related functions */ |
305 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx; |
306 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
307 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
308 |
|
|
309 |
|
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; |
310 |
|
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; |
311 |
|
|
312 |
|
interpolate8x8_avg2 = interpolate8x8_avg2_mmx; |
313 |
|
interpolate8x8_avg4 = interpolate8x8_avg4_mmx; |
314 |
|
|
315 |
/* Image RGB->YV12 related functions */ |
/* Image RGB->YV12 related functions */ |
316 |
rgb24_to_yv12 = rgb24_to_yv12_mmx; |
rgb24_to_yv12 = rgb24_to_yv12_mmx; |
317 |
rgb32_to_yv12 = rgb32_to_yv12_mmx; |
rgb32_to_yv12 = rgb32_to_yv12_mmx; |
329 |
calc_cbp = calc_cbp_mmx; |
calc_cbp = calc_cbp_mmx; |
330 |
sad16 = sad16_mmx; |
sad16 = sad16_mmx; |
331 |
sad8 = sad8_mmx; |
sad8 = sad8_mmx; |
332 |
|
sad16bi = sad16bi_mmx; |
333 |
|
sad8bi = sad8bi_mmx; |
334 |
dev16 = dev16_mmx; |
dev16 = dev16_mmx; |
335 |
|
sad16v = sad16v_mmx; |
336 |
|
|
337 |
} |
} |
338 |
|
|
339 |
|
/* these 3dnow functions are faster than mmx, but slower than xmm. */ |
340 |
|
if ((cpu_flags & XVID_CPU_3DNOW) > 0) { |
341 |
|
|
342 |
|
/* ME functions */ |
343 |
|
sad16bi = sad16bi_3dn; |
344 |
|
sad8bi = sad8bi_3dn; |
345 |
|
} |
346 |
|
|
347 |
|
|
348 |
if ((cpu_flags & XVID_CPU_MMXEXT) > 0) { |
if ((cpu_flags & XVID_CPU_MMXEXT) > 0) { |
349 |
|
|
350 |
/* Inverse DCT */ |
/* Inverse DCT */ |
367 |
|
|
368 |
/* ME functions */ |
/* ME functions */ |
369 |
sad16 = sad16_xmm; |
sad16 = sad16_xmm; |
|
sad16bi = sad16bi_xmm; |
|
370 |
sad8 = sad8_xmm; |
sad8 = sad8_xmm; |
371 |
|
sad16bi = sad16bi_xmm; |
372 |
|
sad8bi = sad8bi_xmm; |
373 |
dev16 = dev16_xmm; |
dev16 = dev16_xmm; |
374 |
|
sad16v = sad16v_xmm; |
375 |
|
fprintf(stderr,"sad16v=XMM\n"); |
376 |
|
|
377 |
} |
} |
378 |
|
|
419 |
sad16bi = sad16bi_ia64; |
sad16bi = sad16bi_ia64; |
420 |
sad8 = sad8_ia64; |
sad8 = sad8_ia64; |
421 |
dev16 = dev16_ia64; |
dev16 = dev16_ia64; |
422 |
|
// Halfpel8_Refine = Halfpel8_Refine_ia64; |
423 |
quant_intra = quant_intra_ia64; |
quant_intra = quant_intra_ia64; |
424 |
dequant_intra = dequant_intra_ia64; |
dequant_intra = dequant_intra_ia64; |
425 |
quant_inter = quant_inter_ia64; |
quant_inter = quant_inter_ia64; |