3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Native API implementation - |
* - Native API implementation - |
5 |
* |
* |
6 |
* Copyright(C) 2001-2003 Peter Ross <pross@xvid.org> |
* Copyright(C) 2001-2004 Peter Ross <pross@xvid.org> |
7 |
* |
* |
8 |
* This program is free software ; you can redistribute it and/or modify |
* 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 |
* it under the terms of the GNU General Public License as published by |
19 |
* along with this program ; if not, write to the Free Software |
* along with this program ; if not, write to the Free Software |
20 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
21 |
* |
* |
22 |
* $Id: xvid.c,v 1.48 2004-03-22 22:36:23 edgomez Exp $ |
* $Id: xvid.c,v 1.58 2004-12-09 23:02:54 edgomez Exp $ |
23 |
* |
* |
24 |
****************************************************************************/ |
****************************************************************************/ |
25 |
|
|
36 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
37 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
38 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
|
#include "image/reduced.h" |
|
39 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
40 |
#include "utils/mbfunctions.h" |
#include "utils/mbfunctions.h" |
41 |
#include "quant/quant.h" |
#include "quant/quant.h" |
51 |
unsigned int xvid_debug = 0; /* xvid debug mask */ |
unsigned int xvid_debug = 0; /* xvid debug mask */ |
52 |
#endif |
#endif |
53 |
|
|
54 |
#if defined(ARCH_IS_IA32) |
#if defined(ARCH_IS_IA32) && defined(_MSC_VER) |
|
#if defined(_MSC_VER) |
|
55 |
# include <windows.h> |
# include <windows.h> |
56 |
#else |
#elif defined(ARCH_IS_IA32) || defined(ARCH_IS_PPC) |
57 |
# include <signal.h> |
# include <signal.h> |
58 |
# include <setjmp.h> |
# include <setjmp.h> |
59 |
|
|
76 |
* 0 : SIGILL was *not* signalled |
* 0 : SIGILL was *not* signalled |
77 |
* 1 : SIGILL was signalled |
* 1 : SIGILL was signalled |
78 |
*/ |
*/ |
79 |
|
#if defined(ARCH_IS_IA32) && defined(_MSC_VER) |
80 |
int |
static int |
81 |
sigill_check(void (*func)()) |
sigill_check(void (*func)()) |
82 |
{ |
{ |
|
#if defined(_MSC_VER) |
|
83 |
_try { |
_try { |
84 |
func(); |
func(); |
85 |
} |
} _except(EXCEPTION_EXECUTE_HANDLER) { |
|
_except(EXCEPTION_EXECUTE_HANDLER) { |
|
86 |
|
|
87 |
if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION) |
if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION) |
88 |
return 1; |
return(1); |
89 |
} |
} |
90 |
return 0; |
return(0); |
91 |
#else |
} |
92 |
|
#elif defined(ARCH_IS_IA32) || defined(ARCH_IS_PPC) |
93 |
|
static int |
94 |
|
sigill_check(void (*func)()) |
95 |
|
{ |
96 |
void * old_handler; |
void * old_handler; |
97 |
int jmpret; |
int jmpret; |
98 |
|
|
99 |
|
/* Set our SIGILL handler */ |
100 |
old_handler = signal(SIGILL, sigill_handler); |
old_handler = signal(SIGILL, sigill_handler); |
101 |
if (old_handler == SIG_ERR) |
|
102 |
{ |
/* Check for error */ |
103 |
return -1; |
if (old_handler == SIG_ERR) { |
104 |
|
return(-1); |
105 |
} |
} |
106 |
|
|
107 |
|
/* Save stack context, so if func triggers a SIGILL, we can still roll |
108 |
|
* back to a valid CPU state */ |
109 |
jmpret = setjmp(mark); |
jmpret = setjmp(mark); |
110 |
if (jmpret == 0) |
|
111 |
{ |
/* If setjmp returned directly, then its returned value is 0, and we still |
112 |
|
* have to test the passed func. Otherwise it means the stack context has |
113 |
|
* been restored by a longjmp() call, which in our case happens only in the |
114 |
|
* signal handler */ |
115 |
|
if (jmpret == 0) { |
116 |
func(); |
func(); |
117 |
} |
} |
118 |
|
|
119 |
|
/* Restore old signal handler */ |
120 |
signal(SIGILL, old_handler); |
signal(SIGILL, old_handler); |
121 |
|
|
122 |
return jmpret; |
return(jmpret); |
|
#endif |
|
123 |
} |
} |
124 |
#endif |
#endif |
125 |
|
|
141 |
#endif |
#endif |
142 |
|
|
143 |
#if defined(ARCH_IS_PPC) |
#if defined(ARCH_IS_PPC) |
144 |
#if defined(ARCH_IS_PPC_ALTIVEC) |
if (!sigill_check(altivec_trigger)) |
145 |
cpu_flags |= XVID_CPU_ALTIVEC; |
cpu_flags |= XVID_CPU_ALTIVEC; |
146 |
#endif |
#endif |
|
#endif |
|
147 |
|
|
148 |
return cpu_flags; |
return cpu_flags; |
149 |
} |
} |
221 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
222 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
223 |
|
|
224 |
|
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_c; |
225 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_c; |
226 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_c; |
227 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_c; |
228 |
|
|
229 |
interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c; |
interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c; |
230 |
interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c; |
interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c; |
231 |
interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c; |
interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c; |
240 |
interpolate8x8_avg2 = interpolate8x8_avg2_c; |
interpolate8x8_avg2 = interpolate8x8_avg2_c; |
241 |
interpolate8x8_avg4 = interpolate8x8_avg4_c; |
interpolate8x8_avg4 = interpolate8x8_avg4_c; |
242 |
|
|
243 |
/* reduced resolution */ |
/* postprocessing */ |
244 |
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_C; |
image_brightness = image_brightness_c; |
|
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_C; |
|
|
vfilter_31 = xvid_VFilter_31_C; |
|
|
hfilter_31 = xvid_HFilter_31_C; |
|
|
filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_C; |
|
|
filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_C; |
|
245 |
|
|
246 |
/* Initialize internal colorspace transformation tables */ |
/* Initialize internal colorspace transformation tables */ |
247 |
colorspace_init(); |
colorspace_init(); |
298 |
dev16 = dev16_c; |
dev16 = dev16_c; |
299 |
sad16v = sad16v_c; |
sad16v = sad16v_c; |
300 |
sse8_16bit = sse8_16bit_c; |
sse8_16bit = sse8_16bit_c; |
301 |
|
sse8_8bit = sse8_8bit_c; |
302 |
|
|
303 |
#if defined(ARCH_IS_IA32) |
#if defined(ARCH_IS_IA32) |
304 |
|
|
|
if ((cpu_flags & XVID_CPU_ASM)) { |
|
|
vfilter_31 = xvid_VFilter_31_x86; |
|
|
hfilter_31 = xvid_HFilter_31_x86; |
|
|
} |
|
|
|
|
305 |
if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || |
if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || |
306 |
(cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || |
(cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || |
307 |
(cpu_flags & XVID_CPU_SSE) || (cpu_flags & XVID_CPU_SSE2)) |
(cpu_flags & XVID_CPU_SSE) || (cpu_flags & XVID_CPU_SSE2)) |
348 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
349 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
350 |
|
|
351 |
|
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_mmx; |
352 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_mmx; |
353 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_mmx; |
354 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_mmx; |
355 |
|
|
356 |
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; |
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; |
357 |
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; |
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; |
358 |
|
|
359 |
interpolate8x8_avg2 = interpolate8x8_avg2_mmx; |
interpolate8x8_avg2 = interpolate8x8_avg2_mmx; |
360 |
interpolate8x8_avg4 = interpolate8x8_avg4_mmx; |
interpolate8x8_avg4 = interpolate8x8_avg4_mmx; |
361 |
|
|
362 |
/* reduced resolution */ |
/* postprocessing */ |
363 |
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_mmx; |
image_brightness = image_brightness_mmx; |
|
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_mmx; |
|
|
hfilter_31 = xvid_HFilter_31_mmx; |
|
|
filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_mmx; |
|
|
filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_mmx; |
|
364 |
|
|
365 |
/* image input xxx_to_yv12 related functions */ |
/* image input xxx_to_yv12 related functions */ |
366 |
yv12_to_yv12 = yv12_to_yv12_mmx; |
yv12_to_yv12 = yv12_to_yv12_mmx; |
387 |
dev16 = dev16_mmx; |
dev16 = dev16_mmx; |
388 |
sad16v = sad16v_mmx; |
sad16v = sad16v_mmx; |
389 |
sse8_16bit = sse8_16bit_mmx; |
sse8_16bit = sse8_16bit_mmx; |
390 |
|
sse8_8bit = sse8_8bit_mmx; |
391 |
} |
} |
392 |
|
|
393 |
/* these 3dnow functions are faster than mmx, but slower than xmm. */ |
/* these 3dnow functions are faster than mmx, but slower than xmm. */ |
415 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
416 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
417 |
|
|
418 |
/* reduced resolution */ |
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_xmm; |
419 |
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_xmm; |
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_xmm; |
420 |
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_xmm; |
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_xmm; |
421 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_xmm; |
422 |
|
|
423 |
/* Quantization */ |
/* Quantization */ |
424 |
quant_mpeg_intra = quant_mpeg_intra_xmm; |
quant_mpeg_intra = quant_mpeg_intra_xmm; |
454 |
|
|
455 |
if ((cpu_flags & XVID_CPU_3DNOWEXT)) { |
if ((cpu_flags & XVID_CPU_3DNOWEXT)) { |
456 |
|
|
|
/* Inverse DCT */ |
|
|
idct = idct_3dne; |
|
|
|
|
457 |
/* Buffer transfer */ |
/* Buffer transfer */ |
458 |
transfer_8to16copy = transfer_8to16copy_3dne; |
transfer_8to16copy = transfer_8to16copy_3dne; |
459 |
transfer_16to8copy = transfer_16to8copy_3dne; |
transfer_16to8copy = transfer_16to8copy_3dne; |
460 |
transfer_8to16sub = transfer_8to16sub_3dne; |
transfer_8to16sub = transfer_8to16sub_3dne; |
461 |
transfer_8to16subro = transfer_8to16subro_3dne; |
transfer_8to16subro = transfer_8to16subro_3dne; |
|
transfer_8to16sub2 = transfer_8to16sub2_3dne; |
|
462 |
transfer_16to8add = transfer_16to8add_3dne; |
transfer_16to8add = transfer_16to8add_3dne; |
463 |
transfer8x8_copy = transfer8x8_copy_3dne; |
transfer8x8_copy = transfer8x8_copy_3dne; |
464 |
|
|
465 |
|
if ((cpu_flags & XVID_CPU_MMXEXT)) { |
466 |
|
/* Inverse DCT */ |
467 |
|
idct = idct_3dne; |
468 |
|
|
469 |
|
/* Buffer transfer */ |
470 |
|
transfer_8to16sub2 = transfer_8to16sub2_3dne; |
471 |
|
|
472 |
|
/* Interpolation */ |
473 |
|
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne; |
474 |
|
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne; |
475 |
|
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne; |
476 |
|
|
477 |
/* Quantization */ |
/* Quantization */ |
478 |
quant_h263_intra = quant_h263_intra_3dne; |
quant_h263_intra = quant_h263_intra_3dne; /* cmov only */ |
479 |
quant_h263_inter = quant_h263_inter_3dne; |
quant_h263_inter = quant_h263_inter_3dne; |
480 |
dequant_mpeg_intra = dequant_mpeg_intra_3dne; |
dequant_mpeg_intra = dequant_mpeg_intra_3dne; /* cmov only */ |
481 |
dequant_mpeg_inter = dequant_mpeg_inter_3dne; |
dequant_mpeg_inter = dequant_mpeg_inter_3dne; |
482 |
dequant_h263_intra = dequant_h263_intra_3dne; |
dequant_h263_intra = dequant_h263_intra_3dne; |
483 |
dequant_h263_inter = dequant_h263_inter_3dne; |
dequant_h263_inter = dequant_h263_inter_3dne; |
484 |
|
|
485 |
/* ME functions */ |
/* ME functions */ |
486 |
calc_cbp = calc_cbp_3dne; |
calc_cbp = calc_cbp_3dne; |
487 |
|
|
488 |
sad16 = sad16_3dne; |
sad16 = sad16_3dne; |
489 |
sad8 = sad8_3dne; |
sad8 = sad8_3dne; |
490 |
sad16bi = sad16bi_3dne; |
sad16bi = sad16bi_3dne; |
491 |
sad8bi = sad8bi_3dne; |
sad8bi = sad8bi_3dne; |
492 |
dev16 = dev16_3dne; |
dev16 = dev16_3dne; |
493 |
|
} |
|
/* Interpolation */ |
|
|
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne; |
|
|
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne; |
|
|
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne; |
|
494 |
} |
} |
495 |
|
|
496 |
if ((cpu_flags & XVID_CPU_SSE2)) { |
if ((cpu_flags & XVID_CPU_SSE2)) { |
510 |
/* DCT operators |
/* DCT operators |
511 |
* no iDCT because it's not "Walken matching" */ |
* no iDCT because it's not "Walken matching" */ |
512 |
fdct = fdct_sse2_skal; |
fdct = fdct_sse2_skal; |
513 |
|
|
514 |
|
/* postprocessing */ |
515 |
|
image_brightness = image_brightness_sse2; |
516 |
} |
} |
517 |
#endif /* ARCH_IS_IA32 */ |
#endif /* ARCH_IS_IA32 */ |
518 |
|
|
543 |
#endif |
#endif |
544 |
|
|
545 |
#if defined(ARCH_IS_PPC) |
#if defined(ARCH_IS_PPC) |
546 |
if ((cpu_flags & XVID_CPU_ASM)) |
if ((cpu_flags & XVID_CPU_ALTIVEC)) { |
547 |
{ |
/* sad operators */ |
548 |
calc_cbp = calc_cbp_ppc; |
sad16 = sad16_altivec_c; |
549 |
} |
sad16bi = sad16bi_altivec_c; |
550 |
|
sad8 = sad8_altivec_c; |
551 |
|
dev16 = dev16_altivec_c; |
552 |
|
|
553 |
|
sse8_16bit = sse8_16bit_altivec_c; |
554 |
|
|
555 |
|
/* mem transfer */ |
556 |
|
transfer_8to16copy = transfer_8to16copy_altivec_c; |
557 |
|
transfer_16to8copy = transfer_16to8copy_altivec_c; |
558 |
|
transfer_8to16sub = transfer_8to16sub_altivec_c; |
559 |
|
transfer_8to16subro = transfer_8to16subro_altivec_c; |
560 |
|
transfer_8to16sub2 = transfer_8to16sub2_altivec_c; |
561 |
|
transfer_16to8add = transfer_16to8add_altivec_c; |
562 |
|
transfer8x8_copy = transfer8x8_copy_altivec_c; |
563 |
|
|
564 |
if ((cpu_flags & XVID_CPU_ALTIVEC)) |
/* Inverse DCT */ |
565 |
{ |
idct = idct_altivec_c; |
566 |
calc_cbp = calc_cbp_altivec; |
|
567 |
fdct = fdct_altivec; |
/* Interpolation */ |
568 |
idct = idct_altivec; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_altivec_c; |
569 |
sadInit = sadInit_altivec; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_altivec_c; |
570 |
sad16 = sad16_altivec; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_altivec_c; |
571 |
sad8 = sad8_altivec; |
|
572 |
dev16 = dev16_altivec; |
interpolate8x8_avg2 = interpolate8x8_avg2_altivec_c; |
573 |
|
interpolate8x8_avg4 = interpolate8x8_avg4_altivec_c; |
574 |
|
|
575 |
|
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_altivec_c; |
576 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_altivec_c; |
577 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_altivec_c; |
578 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_altivec_c; |
579 |
|
|
580 |
|
/* Colorspace conversion */ |
581 |
|
bgra_to_yv12 = bgra_to_yv12_altivec_c; |
582 |
|
abgr_to_yv12 = abgr_to_yv12_altivec_c; |
583 |
|
rgba_to_yv12 = rgba_to_yv12_altivec_c; |
584 |
|
argb_to_yv12 = argb_to_yv12_altivec_c; |
585 |
|
|
586 |
|
yuyv_to_yv12 = yuyv_to_yv12_altivec_c; |
587 |
|
uyvy_to_yv12 = uyvy_to_yv12_altivec_c; |
588 |
|
|
589 |
|
yv12_to_yuyv = yv12_to_yuyv_altivec_c; |
590 |
|
yv12_to_uyvy = yv12_to_uyvy_altivec_c; |
591 |
|
|
592 |
|
/* Quantization */ |
593 |
|
quant_h263_intra = quant_h263_intra_altivec_c; |
594 |
|
quant_h263_inter = quant_h263_inter_altivec_c; |
595 |
|
dequant_h263_intra = dequant_h263_intra_altivec_c; |
596 |
|
dequant_h263_inter = dequant_h263_inter_altivec_c; |
597 |
|
|
598 |
|
/* Qpel stuff */ |
599 |
|
xvid_QP_Funcs = &xvid_QP_Funcs_Altivec_C; |
600 |
|
xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_Altivec_C; |
601 |
} |
} |
602 |
#endif |
#endif |
603 |
|
|
605 |
xvid_debug = init->debug; |
xvid_debug = init->debug; |
606 |
#endif |
#endif |
607 |
|
|
608 |
return 0; |
return(0); |
609 |
} |
} |
610 |
|
|
611 |
|
|
616 |
return XVID_ERR_VERSION; |
return XVID_ERR_VERSION; |
617 |
|
|
618 |
info->actual_version = XVID_VERSION; |
info->actual_version = XVID_VERSION; |
619 |
info->build = "xvid-1.0.0"; |
info->build = "xvid-1.1-cvshead"; |
620 |
info->cpu_flags = detect_cpu_flags(); |
info->cpu_flags = detect_cpu_flags(); |
621 |
|
|
622 |
#if defined(_SMP) && defined(WIN32) |
#if defined(_SMP) && defined(WIN32) |