3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Native API implementation - |
* - Native API implementation - |
5 |
* |
* |
6 |
|
* Copyright(C) 2001-2002 Peter Ross <pross@xvid.org> |
7 |
|
* |
8 |
* This program is an implementation of a part of one or more MPEG-4 |
* This program is an implementation of a part of one or more MPEG-4 |
9 |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
10 |
* to use this software module in hardware or software products are |
* to use this software module in hardware or software products are |
28 |
* along with this program ; if not, write to the Free Software |
* along with this program ; if not, write to the Free Software |
29 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
30 |
* |
* |
31 |
****************************************************************************/ |
* $Id: xvid.c,v 1.37 2002-09-21 03:07:56 suxen_drol Exp $ |
|
|
|
|
/***************************************************************************** |
|
|
* |
|
|
* History |
|
|
* |
|
|
* - 23.06.2002 added XVID_CPU_CHKONLY |
|
|
* - 17.03.2002 Added interpolate8x8_halfpel_hv_xmm |
|
|
* - 22.12.2001 API change: added xvid_init() - Isibaar |
|
|
* - 16.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
|
|
* |
|
|
* $Id: xvid.c,v 1.27 2002-07-09 01:37:21 chenm001 Exp $ |
|
32 |
* |
* |
33 |
****************************************************************************/ |
****************************************************************************/ |
34 |
|
|
43 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
44 |
#include "quant/quant_h263.h" |
#include "quant/quant_h263.h" |
45 |
#include "quant/quant_mpeg4.h" |
#include "quant/quant_mpeg4.h" |
46 |
|
#include "motion/motion.h" |
47 |
#include "motion/sad.h" |
#include "motion/sad.h" |
48 |
#include "utils/emms.h" |
#include "utils/emms.h" |
49 |
#include "utils/timer.h" |
#include "utils/timer.h" |
50 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
51 |
|
|
52 |
|
#if defined(ARCH_X86) && defined(EXPERIMENTAL_SSE2_CODE) |
53 |
|
|
54 |
|
#ifdef WIN32 |
55 |
|
#include <windows.h> |
56 |
|
#else |
57 |
|
#include <signal.h> |
58 |
|
#include <setjmp.h> |
59 |
|
#endif |
60 |
|
|
61 |
|
|
62 |
|
#ifndef WIN32 |
63 |
|
|
64 |
|
static jmp_buf mark; |
65 |
|
|
66 |
|
static void |
67 |
|
sigill_handler(int signal) |
68 |
|
{ |
69 |
|
longjmp(mark, 1); |
70 |
|
} |
71 |
|
#endif |
72 |
|
|
73 |
|
|
74 |
|
/* |
75 |
|
* Calls the funcptr, and returns whether SIGILL (illegal instruction) was signalled |
76 |
|
* Return values: |
77 |
|
* -1 : could not determine |
78 |
|
* 0 : SIGILL was *not* signalled |
79 |
|
* 1 : SIGILL was signalled |
80 |
|
*/ |
81 |
|
|
82 |
|
int |
83 |
|
sigill_check(void (*func)()) |
84 |
|
{ |
85 |
|
#ifdef WIN32 |
86 |
|
_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 |
/***************************************************************************** |
/***************************************************************************** |
120 |
* XviD Init Entry point |
* XviD Init Entry point |
121 |
* |
* |
147 |
/* Inform the client the core build - unused because we're still alpha */ |
/* Inform the client the core build - unused because we're still alpha */ |
148 |
init_param->core_build = 1000; |
init_param->core_build = 1000; |
149 |
|
|
|
if ((init_param->cpu_flags & XVID_CPU_CHKONLY)) |
|
|
{ |
|
|
init_param->cpu_flags = check_cpu_features(); |
|
|
return XVID_ERR_OK; |
|
|
} |
|
|
|
|
150 |
/* Do we have to force CPU features ? */ |
/* Do we have to force CPU features ? */ |
151 |
if ((init_param->cpu_flags & XVID_CPU_FORCE) > 0) { |
if ((init_param->cpu_flags & XVID_CPU_FORCE)) { |
152 |
|
|
153 |
cpu_flags = init_param->cpu_flags; |
cpu_flags = init_param->cpu_flags; |
154 |
|
|
155 |
} else { |
} else { |
156 |
|
|
157 |
cpu_flags = check_cpu_features(); |
cpu_flags = check_cpu_features(); |
158 |
|
|
159 |
|
#if defined(ARCH_X86) && defined(EXPERIMENTAL_SSE2_CODE) |
160 |
|
if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger)) |
161 |
|
cpu_flags &= ~XVID_CPU_SSE; |
162 |
|
|
163 |
|
if ((cpu_flags & XVID_CPU_SSE2) && sigill_check(sse2_os_trigger)) |
164 |
|
cpu_flags &= ~XVID_CPU_SSE2; |
165 |
|
#endif |
166 |
|
} |
167 |
|
|
168 |
|
if ((init_param->cpu_flags & XVID_CPU_CHKONLY)) |
169 |
|
{ |
170 |
init_param->cpu_flags = cpu_flags; |
init_param->cpu_flags = cpu_flags; |
171 |
|
return XVID_ERR_OK; |
172 |
} |
} |
173 |
|
|
174 |
|
init_param->cpu_flags = cpu_flags; |
175 |
|
|
176 |
|
|
177 |
/* Initialize the function pointers */ |
/* Initialize the function pointers */ |
178 |
idct_int32_init(); |
idct_int32_init(); |
179 |
init_vlc_tables(); |
init_vlc_tables(); |
236 |
/* Functions used in motion estimation algorithms */ |
/* Functions used in motion estimation algorithms */ |
237 |
calc_cbp = calc_cbp_c; |
calc_cbp = calc_cbp_c; |
238 |
sad16 = sad16_c; |
sad16 = sad16_c; |
|
sad16bi = sad16bi_c; |
|
239 |
sad8 = sad8_c; |
sad8 = sad8_c; |
240 |
|
sad16bi = sad16bi_c; |
241 |
|
sad8bi = sad8bi_c; |
242 |
dev16 = dev16_c; |
dev16 = dev16_c; |
243 |
|
|
244 |
|
Halfpel8_Refine = Halfpel8_Refine_c; |
245 |
|
|
246 |
#ifdef ARCH_X86 |
#ifdef ARCH_X86 |
247 |
if ((cpu_flags & XVID_CPU_MMX) > 0) { |
if ((cpu_flags & XVID_CPU_MMX) > 0) { |
248 |
|
|
295 |
calc_cbp = calc_cbp_mmx; |
calc_cbp = calc_cbp_mmx; |
296 |
sad16 = sad16_mmx; |
sad16 = sad16_mmx; |
297 |
sad8 = sad8_mmx; |
sad8 = sad8_mmx; |
298 |
|
sad16bi = sad16bi_mmx; |
299 |
|
sad8bi = sad8bi_mmx; |
300 |
dev16 = dev16_mmx; |
dev16 = dev16_mmx; |
301 |
|
|
302 |
} |
} |
303 |
|
|
304 |
|
/* these 3dnow functions are faster than mmx, but slower than xmm. */ |
305 |
|
if ((cpu_flags & XVID_CPU_3DNOW) > 0) { |
306 |
|
|
307 |
|
/* ME functions */ |
308 |
|
sad16bi = sad16bi_3dn; |
309 |
|
sad8bi = sad8bi_3dn; |
310 |
|
} |
311 |
|
|
312 |
|
|
313 |
if ((cpu_flags & XVID_CPU_MMXEXT) > 0) { |
if ((cpu_flags & XVID_CPU_MMXEXT) > 0) { |
314 |
|
|
315 |
/* Inverse DCT */ |
/* Inverse DCT */ |
332 |
|
|
333 |
/* ME functions */ |
/* ME functions */ |
334 |
sad16 = sad16_xmm; |
sad16 = sad16_xmm; |
|
sad16bi = sad16bi_xmm; |
|
335 |
sad8 = sad8_xmm; |
sad8 = sad8_xmm; |
336 |
|
sad16bi = sad16bi_xmm; |
337 |
|
sad8bi = sad8bi_xmm; |
338 |
dev16 = dev16_xmm; |
dev16 = dev16_xmm; |
339 |
|
|
340 |
} |
} |
382 |
sad16bi = sad16bi_ia64; |
sad16bi = sad16bi_ia64; |
383 |
sad8 = sad8_ia64; |
sad8 = sad8_ia64; |
384 |
dev16 = dev16_ia64; |
dev16 = dev16_ia64; |
385 |
|
Halfpel8_Refine = Halfpel8_Refine_ia64; |
386 |
quant_intra = quant_intra_ia64; |
quant_intra = quant_intra_ia64; |
387 |
dequant_intra = dequant_intra_ia64; |
dequant_intra = dequant_intra_ia64; |
388 |
quant_inter = quant_inter_ia64; |
quant_inter = quant_inter_ia64; |
464 |
{ |
{ |
465 |
switch (opt) { |
switch (opt) { |
466 |
case XVID_ENC_ENCODE: |
case XVID_ENC_ENCODE: |
|
#ifdef BFRAMES |
|
|
if (((Encoder *) handle)->mbParam.max_bframes >= 0) |
|
|
return encoder_encode_bframes((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
|
|
(XVID_ENC_STATS *) param2); |
|
|
else |
|
|
#endif |
|
467 |
return encoder_encode((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
return encoder_encode((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
468 |
(XVID_ENC_STATS *) param2); |
(XVID_ENC_STATS *) param2); |
469 |
|
|