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 1.51 2004-04-12 14:05:08 edgomez Exp $ |
* $Id: xvid.c,v 1.60 2005-01-05 23:02:15 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) && defined(_MSC_VER) |
#if (defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64)) && defined(_MSC_VER) |
55 |
# include <windows.h> |
# include <windows.h> |
56 |
#elif defined(ARCH_IS_IA32) || defined(ARCH_IS_PPC) |
#elif defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64) || 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) |
#if (defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64)) && defined(_MSC_VER) |
80 |
static int |
static int |
81 |
sigill_check(void (*func)()) |
sigill_check(void (*func)()) |
82 |
{ |
{ |
89 |
} |
} |
90 |
return(0); |
return(0); |
91 |
} |
} |
92 |
#elif defined(ARCH_IS_IA32) || defined(ARCH_IS_PPC) |
#elif defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64) || defined(ARCH_IS_PPC) |
93 |
static int |
static int |
94 |
sigill_check(void (*func)()) |
sigill_check(void (*func)()) |
95 |
{ |
{ |
131 |
/* enable native assembly optimizations by default */ |
/* enable native assembly optimizations by default */ |
132 |
unsigned int cpu_flags = XVID_CPU_ASM; |
unsigned int cpu_flags = XVID_CPU_ASM; |
133 |
|
|
134 |
#if defined(ARCH_IS_IA32) |
#if defined(ARCH_IS_IA32) || defined(ARCH_IS_X86_64) |
135 |
cpu_flags |= check_cpu_features(); |
cpu_flags |= check_cpu_features(); |
136 |
if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger)) |
if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger)) |
137 |
cpu_flags &= ~XVID_CPU_SSE; |
cpu_flags &= ~XVID_CPU_SSE; |
210 |
transfer_8to16sub = transfer_8to16sub_c; |
transfer_8to16sub = transfer_8to16sub_c; |
211 |
transfer_8to16subro = transfer_8to16subro_c; |
transfer_8to16subro = transfer_8to16subro_c; |
212 |
transfer_8to16sub2 = transfer_8to16sub2_c; |
transfer_8to16sub2 = transfer_8to16sub2_c; |
213 |
|
transfer_8to16sub2ro = transfer_8to16sub2ro_c; |
214 |
transfer_16to8add = transfer_16to8add_c; |
transfer_16to8add = transfer_16to8add_c; |
215 |
transfer8x8_copy = transfer8x8_copy_c; |
transfer8x8_copy = transfer8x8_copy_c; |
216 |
|
|
222 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
223 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
224 |
|
|
225 |
|
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_c; |
226 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_c; |
227 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_c; |
228 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_c; |
229 |
|
|
230 |
interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c; |
interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c; |
231 |
interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c; |
interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c; |
232 |
interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c; |
interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c; |
244 |
/* postprocessing */ |
/* postprocessing */ |
245 |
image_brightness = image_brightness_c; |
image_brightness = image_brightness_c; |
246 |
|
|
|
/* reduced resolution */ |
|
|
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_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; |
|
|
|
|
247 |
/* Initialize internal colorspace transformation tables */ |
/* Initialize internal colorspace transformation tables */ |
248 |
colorspace_init(); |
colorspace_init(); |
249 |
|
|
299 |
dev16 = dev16_c; |
dev16 = dev16_c; |
300 |
sad16v = sad16v_c; |
sad16v = sad16v_c; |
301 |
sse8_16bit = sse8_16bit_c; |
sse8_16bit = sse8_16bit_c; |
302 |
|
sse8_8bit = sse8_8bit_c; |
303 |
|
|
304 |
#if defined(ARCH_IS_IA32) |
#if defined(ARCH_IS_IA32) |
305 |
|
|
|
if ((cpu_flags & XVID_CPU_ASM)) { |
|
|
vfilter_31 = xvid_VFilter_31_x86; |
|
|
hfilter_31 = xvid_HFilter_31_x86; |
|
|
} |
|
|
|
|
306 |
if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || |
if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || |
307 |
(cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || |
(cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || |
308 |
(cpu_flags & XVID_CPU_SSE) || (cpu_flags & XVID_CPU_SSE2)) |
(cpu_flags & XVID_CPU_SSE) || (cpu_flags & XVID_CPU_SSE2)) |
349 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
350 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
351 |
|
|
352 |
|
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_mmx; |
353 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_mmx; |
354 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_mmx; |
355 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_mmx; |
356 |
|
|
357 |
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; |
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; |
358 |
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; |
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; |
359 |
|
|
363 |
/* postprocessing */ |
/* postprocessing */ |
364 |
image_brightness = image_brightness_mmx; |
image_brightness = image_brightness_mmx; |
365 |
|
|
|
/* reduced resolution */ |
|
|
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_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; |
|
|
|
|
366 |
/* image input xxx_to_yv12 related functions */ |
/* image input xxx_to_yv12 related functions */ |
367 |
yv12_to_yv12 = yv12_to_yv12_mmx; |
yv12_to_yv12 = yv12_to_yv12_mmx; |
368 |
bgr_to_yv12 = bgr_to_yv12_mmx; |
bgr_to_yv12 = bgr_to_yv12_mmx; |
388 |
dev16 = dev16_mmx; |
dev16 = dev16_mmx; |
389 |
sad16v = sad16v_mmx; |
sad16v = sad16v_mmx; |
390 |
sse8_16bit = sse8_16bit_mmx; |
sse8_16bit = sse8_16bit_mmx; |
391 |
|
sse8_8bit = sse8_8bit_mmx; |
392 |
} |
} |
393 |
|
|
394 |
/* these 3dnow functions are faster than mmx, but slower than xmm. */ |
/* these 3dnow functions are faster than mmx, but slower than xmm. */ |
416 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
417 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
418 |
|
|
419 |
/* reduced resolution */ |
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_xmm; |
420 |
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_xmm; |
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_xmm; |
421 |
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_xmm; |
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_xmm; |
422 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_xmm; |
423 |
|
|
424 |
/* Quantization */ |
/* Quantization */ |
425 |
quant_mpeg_intra = quant_mpeg_intra_xmm; |
quant_mpeg_intra = quant_mpeg_intra_xmm; |
430 |
|
|
431 |
/* Buffer transfer */ |
/* Buffer transfer */ |
432 |
transfer_8to16sub2 = transfer_8to16sub2_xmm; |
transfer_8to16sub2 = transfer_8to16sub2_xmm; |
433 |
|
transfer_8to16sub2ro = transfer_8to16sub2ro_xmm; |
434 |
|
|
435 |
/* Colorspace transformation */ |
/* Colorspace transformation */ |
436 |
yv12_to_yv12 = yv12_to_yv12_xmm; |
yv12_to_yv12 = yv12_to_yv12_xmm; |
512 |
/* DCT operators |
/* DCT operators |
513 |
* no iDCT because it's not "Walken matching" */ |
* no iDCT because it's not "Walken matching" */ |
514 |
fdct = fdct_sse2_skal; |
fdct = fdct_sse2_skal; |
515 |
|
|
516 |
|
/* postprocessing */ |
517 |
|
image_brightness = image_brightness_sse2; |
518 |
} |
} |
519 |
#endif /* ARCH_IS_IA32 */ |
#endif /* ARCH_IS_IA32 */ |
520 |
|
|
574 |
interpolate8x8_avg2 = interpolate8x8_avg2_altivec_c; |
interpolate8x8_avg2 = interpolate8x8_avg2_altivec_c; |
575 |
interpolate8x8_avg4 = interpolate8x8_avg4_altivec_c; |
interpolate8x8_avg4 = interpolate8x8_avg4_altivec_c; |
576 |
|
|
577 |
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_altivec_c; |
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_altivec_c; |
578 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_altivec_c; |
579 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_altivec_c; |
580 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_altivec_c; |
581 |
|
|
582 |
/* Colorspace conversion */ |
/* Colorspace conversion */ |
583 |
bgra_to_yv12 = bgra_to_yv12_altivec_c; |
bgra_to_yv12 = bgra_to_yv12_altivec_c; |
596 |
quant_h263_inter = quant_h263_inter_altivec_c; |
quant_h263_inter = quant_h263_inter_altivec_c; |
597 |
dequant_h263_intra = dequant_h263_intra_altivec_c; |
dequant_h263_intra = dequant_h263_intra_altivec_c; |
598 |
dequant_h263_inter = dequant_h263_inter_altivec_c; |
dequant_h263_inter = dequant_h263_inter_altivec_c; |
599 |
|
|
600 |
|
/* Qpel stuff */ |
601 |
|
xvid_QP_Funcs = &xvid_QP_Funcs_Altivec_C; |
602 |
|
xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_Altivec_C; |
603 |
|
} |
604 |
|
#endif |
605 |
|
|
606 |
|
#if defined(ARCH_IS_X86_64) |
607 |
|
/* For now, only XVID_CPU_ASM is looked for, so user can still |
608 |
|
* disable asm usage the usual way. When Intel EMT64 cpus will |
609 |
|
* be out, maybe we'll have to check more precisely what cpu |
610 |
|
* features there really are. */ |
611 |
|
if (cpu_flags & XVID_CPU_ASM) { |
612 |
|
/* SIMD state flusher */ |
613 |
|
emms = emms_3dn; |
614 |
|
|
615 |
|
/* DCT operators */ |
616 |
|
fdct = fdct_skal_x86_64; |
617 |
|
idct = idct_x86_64; |
618 |
|
|
619 |
|
/* SAD operators */ |
620 |
|
sad16 = sad16_x86_64; |
621 |
|
sad8 = sad8_x86_64; |
622 |
|
sad16bi = sad16bi_x86_64; |
623 |
|
sad8bi = sad8bi_x86_64; |
624 |
|
dev16 = dev16_x86_64; |
625 |
|
sad16v = sad16v_x86_64; |
626 |
|
sse8_16bit = sse8_16bit_x86_64; |
627 |
|
sse8_8bit = sse8_8bit_x86_64; |
628 |
|
|
629 |
|
/* Interpolation operators */ |
630 |
|
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_x86_64; |
631 |
|
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_x86_64; |
632 |
|
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_x86_64; |
633 |
|
|
634 |
|
interpolate8x8_halfpel_add = interpolate8x8_halfpel_add_x86_64; |
635 |
|
interpolate8x8_halfpel_h_add = interpolate8x8_halfpel_h_add_x86_64; |
636 |
|
interpolate8x8_halfpel_v_add = interpolate8x8_halfpel_v_add_x86_64; |
637 |
|
interpolate8x8_halfpel_hv_add = interpolate8x8_halfpel_hv_add_x86_64; |
638 |
|
|
639 |
|
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_x86_64; |
640 |
|
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_x86_64; |
641 |
|
|
642 |
|
interpolate8x8_avg2 = interpolate8x8_avg2_x86_64; |
643 |
|
interpolate8x8_avg4 = interpolate8x8_avg4_x86_64; |
644 |
|
|
645 |
|
/* Quantization related functions */ |
646 |
|
quant_h263_intra = quant_h263_intra_x86_64; |
647 |
|
quant_h263_inter = quant_h263_inter_x86_64; |
648 |
|
dequant_h263_intra = dequant_h263_intra_x86_64; |
649 |
|
dequant_h263_inter = dequant_h263_inter_x86_64; |
650 |
|
quant_mpeg_intra = quant_mpeg_intra_x86_64; |
651 |
|
quant_mpeg_inter = quant_mpeg_inter_x86_64; |
652 |
|
dequant_mpeg_intra = dequant_mpeg_intra_x86_64; |
653 |
|
dequant_mpeg_inter = dequant_mpeg_inter_x86_64; |
654 |
|
|
655 |
|
/* Block related functions */ |
656 |
|
transfer_8to16copy = transfer_8to16copy_x86_64; |
657 |
|
transfer_16to8copy = transfer_16to8copy_x86_64; |
658 |
|
transfer_8to16sub = transfer_8to16sub_x86_64; |
659 |
|
transfer_8to16subro = transfer_8to16subro_x86_64; |
660 |
|
transfer_8to16sub2 = transfer_8to16sub2_x86_64; |
661 |
|
transfer_8to16sub2ro= transfer_8to16sub2ro_x86_64; |
662 |
|
transfer_16to8add = transfer_16to8add_x86_64; |
663 |
|
transfer8x8_copy = transfer8x8_copy_x86_64; |
664 |
|
|
665 |
|
/* Qpel stuff */ |
666 |
|
xvid_QP_Funcs = &xvid_QP_Funcs_x86_64; |
667 |
|
xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_x86_64; |
668 |
|
|
669 |
|
/* Interlacing Functions */ |
670 |
|
MBFieldTest = MBFieldTest_x86_64; |
671 |
} |
} |
672 |
#endif |
#endif |
673 |
|
|
686 |
return XVID_ERR_VERSION; |
return XVID_ERR_VERSION; |
687 |
|
|
688 |
info->actual_version = XVID_VERSION; |
info->actual_version = XVID_VERSION; |
689 |
info->build = "xvid-1.0.0"; |
info->build = "xvid-1.1-cvshead"; |
690 |
info->cpu_flags = detect_cpu_flags(); |
info->cpu_flags = detect_cpu_flags(); |
691 |
|
|
692 |
#if defined(_SMP) && defined(WIN32) |
#if defined(_SMP) && defined(WIN32) |