27 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
28 |
* |
* |
29 |
****************************************************************************/ |
****************************************************************************/ |
30 |
|
|
31 |
/***************************************************************************** |
/***************************************************************************** |
32 |
* |
* |
33 |
* History |
* History |
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 1.22 2002-06-23 19:48:06 edgomez Exp $ |
* $Id: xvid.c,v 1.33.2.3 2002-09-25 16:51:05 h 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 |
|
|
|
#ifdef ARCH_X86 |
|
167 |
cpu_flags = check_cpu_features(); |
cpu_flags = check_cpu_features(); |
168 |
#else |
|
169 |
cpu_flags = 0; |
#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 |
#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; |
249 |
/* Functions used in motion estimation algorithms */ |
/* Functions used in motion estimation algorithms */ |
250 |
calc_cbp = calc_cbp_c; |
calc_cbp = calc_cbp_c; |
251 |
sad16 = sad16_c; |
sad16 = sad16_c; |
|
sad16bi = sad16bi_c; |
|
252 |
sad8 = sad8_c; |
sad8 = sad8_c; |
253 |
|
sad16bi = sad16bi_c; |
254 |
|
sad8bi = sad8bi_c; |
255 |
dev16 = dev16_c; |
dev16 = dev16_c; |
256 |
|
sad16v = sad16v_c; |
257 |
|
|
258 |
|
// Halfpel8_Refine = Halfpel8_Refine_c; |
259 |
|
|
260 |
#ifdef ARCH_X86 |
#ifdef ARCH_X86 |
261 |
if ((cpu_flags & XVID_CPU_MMX) > 0) { |
if ((cpu_flags & XVID_CPU_MMX) > 0) { |
286 |
transfer_16to8add = transfer_16to8add_mmx; |
transfer_16to8add = transfer_16to8add_mmx; |
287 |
transfer8x8_copy = transfer8x8_copy_mmx; |
transfer8x8_copy = transfer8x8_copy_mmx; |
288 |
|
|
289 |
|
/* Interlacing functions */ |
290 |
|
MBFieldTest = MBFieldTest_mmx; |
291 |
|
|
292 |
/* Image Interpolation related functions */ |
/* Image Interpolation related functions */ |
293 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx; |
311 |
calc_cbp = calc_cbp_mmx; |
calc_cbp = calc_cbp_mmx; |
312 |
sad16 = sad16_mmx; |
sad16 = sad16_mmx; |
313 |
sad8 = sad8_mmx; |
sad8 = sad8_mmx; |
314 |
|
sad16bi = sad16bi_mmx; |
315 |
|
sad8bi = sad8bi_mmx; |
316 |
dev16 = dev16_mmx; |
dev16 = dev16_mmx; |
317 |
|
sad16v = sad16v_mmx; |
318 |
|
|
319 |
|
} |
320 |
|
|
321 |
|
/* these 3dnow functions are faster than mmx, but slower than xmm. */ |
322 |
|
if ((cpu_flags & XVID_CPU_3DNOW) > 0) { |
323 |
|
|
324 |
|
/* ME functions */ |
325 |
|
sad16bi = sad16bi_3dn; |
326 |
|
sad8bi = sad8bi_3dn; |
327 |
} |
} |
328 |
|
|
329 |
|
|
330 |
if ((cpu_flags & XVID_CPU_MMXEXT) > 0) { |
if ((cpu_flags & XVID_CPU_MMXEXT) > 0) { |
331 |
|
|
332 |
/* Inverse DCT */ |
/* Inverse DCT */ |
337 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
338 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
339 |
|
|
340 |
|
/* Quantization */ |
341 |
|
dequant_intra = dequant_intra_xmm; |
342 |
|
dequant_inter = dequant_inter_xmm; |
343 |
|
|
344 |
/* Buffer transfer */ |
/* Buffer transfer */ |
345 |
transfer_8to16sub2 = transfer_8to16sub2_xmm; |
transfer_8to16sub2 = transfer_8to16sub2_xmm; |
346 |
|
|
350 |
/* ME functions */ |
/* ME functions */ |
351 |
sad16 = sad16_xmm; |
sad16 = sad16_xmm; |
352 |
sad8 = sad8_xmm; |
sad8 = sad8_xmm; |
353 |
|
sad16bi = sad16bi_xmm; |
354 |
|
sad8bi = sad8bi_xmm; |
355 |
dev16 = dev16_xmm; |
dev16 = dev16_xmm; |
356 |
|
sad16v = sad16v_xmm; |
357 |
|
fprintf(stderr,"sad16v=XMM\n"); |
358 |
|
|
359 |
} |
} |
360 |
|
|
369 |
if ((cpu_flags & XVID_CPU_SSE2) > 0) { |
if ((cpu_flags & XVID_CPU_SSE2) > 0) { |
370 |
#ifdef EXPERIMENTAL_SSE2_CODE |
#ifdef EXPERIMENTAL_SSE2_CODE |
371 |
|
|
372 |
|
calc_cbp = calc_cbp_sse2; |
373 |
|
|
374 |
/* Quantization */ |
/* Quantization */ |
375 |
quant_intra = quant_intra_sse2; |
quant_intra = quant_intra_sse2; |
376 |
dequant_intra = dequant_intra_sse2; |
dequant_intra = dequant_intra_sse2; |
378 |
dequant_inter = dequant_inter_sse2; |
dequant_inter = dequant_inter_sse2; |
379 |
|
|
380 |
/* ME */ |
/* ME */ |
|
calc_cbp = calc_cbp_sse2; |
|
381 |
sad16 = sad16_sse2; |
sad16 = sad16_sse2; |
382 |
dev16 = dev16_sse2; |
dev16 = dev16_sse2; |
383 |
|
|
401 |
sad16bi = sad16bi_ia64; |
sad16bi = sad16bi_ia64; |
402 |
sad8 = sad8_ia64; |
sad8 = sad8_ia64; |
403 |
dev16 = dev16_ia64; |
dev16 = dev16_ia64; |
404 |
|
// Halfpel8_Refine = Halfpel8_Refine_ia64; |
405 |
quant_intra = quant_intra_ia64; |
quant_intra = quant_intra_ia64; |
406 |
dequant_intra = dequant_intra_ia64; |
dequant_intra = dequant_intra_ia64; |
407 |
quant_inter = quant_inter_ia64; |
quant_inter = quant_inter_ia64; |