237 |
|
|
238 |
/* reduced resoltuion */ |
/* reduced resoltuion */ |
239 |
|
|
240 |
|
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_C; |
241 |
|
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_C; |
242 |
#ifdef ARCH_X86 |
#ifdef ARCH_X86 |
243 |
vfilter_31 = xvid_VFilter_31_x86; |
vfilter_31 = xvid_VFilter_31_x86; |
244 |
hfilter_31 = xvid_HFilter_31_x86; |
hfilter_31 = xvid_HFilter_31_x86; |
245 |
#else |
#else |
|
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_C; |
|
|
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_C; |
|
246 |
vfilter_31 = xvid_VFilter_31_C; |
vfilter_31 = xvid_VFilter_31_C; |
247 |
hfilter_31 = xvid_HFilter_31_C; |
hfilter_31 = xvid_HFilter_31_C; |
248 |
#endif |
#endif |
249 |
|
filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_C; |
250 |
|
filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_C; |
251 |
|
|
252 |
/* Initialize internal colorspace transformation tables */ |
/* Initialize internal colorspace transformation tables */ |
253 |
colorspace_init(); |
colorspace_init(); |
356 |
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_mmx; |
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_mmx; |
357 |
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_mmx; |
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_mmx; |
358 |
hfilter_31 = xvid_HFilter_31_mmx; |
hfilter_31 = xvid_HFilter_31_mmx; |
359 |
|
filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_mmx; |
360 |
|
filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_mmx; |
361 |
|
|
362 |
/* image input xxx_to_yv12 related functions */ |
/* image input xxx_to_yv12 related functions */ |
363 |
yv12_to_yv12 = yv12_to_yv12_mmx; |
yv12_to_yv12 = yv12_to_yv12_mmx; |
413 |
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_xmm; |
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_xmm; |
414 |
|
|
415 |
/* Quantization */ |
/* Quantization */ |
416 |
|
quant4_intra = quant4_intra_xmm; |
417 |
|
quant4_inter = quant4_inter_xmm; |
418 |
|
|
419 |
dequant_intra = dequant_intra_xmm; |
dequant_intra = dequant_intra_xmm; |
420 |
dequant_inter = dequant_inter_xmm; |
dequant_inter = dequant_inter_xmm; |
421 |
|
|
444 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn; |
445 |
} |
} |
446 |
|
|
447 |
|
if ((cpu_flags & XVID_CPU_3DNOWEXT) > 0) { |
448 |
|
|
449 |
|
/* Inverse DCT */ |
450 |
|
idct = idct_3dne; |
451 |
|
|
452 |
|
/* Buffer transfer */ |
453 |
|
transfer_8to16copy = transfer_8to16copy_3dne; |
454 |
|
transfer_16to8copy = transfer_16to8copy_3dne; |
455 |
|
transfer_8to16sub = transfer_8to16sub_3dne; |
456 |
|
transfer_8to16sub2 = transfer_8to16sub2_3dne; |
457 |
|
transfer_16to8add = transfer_16to8add_3dne; |
458 |
|
transfer8x8_copy = transfer8x8_copy_3dne; |
459 |
|
|
460 |
|
/* Quantization */ |
461 |
|
dequant4_intra = dequant4_intra_3dne; |
462 |
|
dequant4_inter = dequant4_inter_3dne; |
463 |
|
quant_intra = quant_intra_3dne; |
464 |
|
quant_inter = quant_inter_3dne; |
465 |
|
dequant_intra = dequant_intra_3dne; |
466 |
|
dequant_inter = dequant_inter_3dne; |
467 |
|
|
468 |
|
/* ME functions */ |
469 |
|
calc_cbp = calc_cbp_3dne; |
470 |
|
sad16 = sad16_3dne; |
471 |
|
sad8 = sad8_3dne; |
472 |
|
sad16bi = sad16bi_3dne; |
473 |
|
sad8bi = sad8bi_3dne; |
474 |
|
dev16 = dev16_3dne; |
475 |
|
|
476 |
|
/* Interpolation */ |
477 |
|
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne; |
478 |
|
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne; |
479 |
|
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne; |
480 |
|
} |
481 |
|
|
482 |
|
|
483 |
if ((cpu_flags & XVID_CPU_SSE2) > 0) { |
if ((cpu_flags & XVID_CPU_SSE2) > 0) { |
484 |
#ifdef EXPERIMENTAL_SSE2_CODE |
#ifdef EXPERIMENTAL_SSE2_CODE |
485 |
|
|
552 |
static int |
static int |
553 |
xvid_init_convert(XVID_INIT_CONVERTINFO* convert) |
xvid_init_convert(XVID_INIT_CONVERTINFO* convert) |
554 |
{ |
{ |
555 |
const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP); |
// const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP); |
556 |
const int width = convert->width; |
const int width = convert->width; |
557 |
const int height = convert->height; |
const int height = convert->height; |
558 |
const int width2 = convert->width/2; |
const int width2 = convert->width/2; |
580 |
} |
} |
581 |
|
|
582 |
|
|
583 |
|
|
584 |
|
void fill8(uint8_t * block, int size, int value) |
585 |
|
{ |
586 |
|
int i; |
587 |
|
for (i = 0; i < size; i++) |
588 |
|
block[i] = value; |
589 |
|
} |
590 |
|
|
591 |
|
void fill16(int16_t * block, int size, int value) |
592 |
|
{ |
593 |
|
int i; |
594 |
|
for (i = 0; i < size; i++) |
595 |
|
block[i] = value; |
596 |
|
} |
597 |
|
|
598 |
|
#define RANDOM(min,max) min + (rand() % (max-min)) |
599 |
|
|
600 |
|
void random8(uint8_t * block, int size, int min, int max) |
601 |
|
{ |
602 |
|
int i; |
603 |
|
for (i = 0; i < size; i++) |
604 |
|
block[i] = RANDOM(min,max); |
605 |
|
} |
606 |
|
|
607 |
|
void random16(int16_t * block, int size, int min, int max) |
608 |
|
{ |
609 |
|
int i; |
610 |
|
for (i = 0; i < size; i++) |
611 |
|
block[i] = RANDOM(min,max); |
612 |
|
} |
613 |
|
|
614 |
|
int compare16(const int16_t * blockA, const int16_t * blockB, int size) |
615 |
|
{ |
616 |
|
int i; |
617 |
|
for (i = 0; i < size; i++) |
618 |
|
if (blockA[i] != blockB[i]) |
619 |
|
return 1; |
620 |
|
|
621 |
|
return 0; |
622 |
|
} |
623 |
|
|
624 |
|
|
625 |
|
|
626 |
|
int test_h263_intra(quanth263_intraFunc * funcA, quanth263_intraFunc * funcB, |
627 |
|
const char * nameA, const char * nameB, |
628 |
|
int min, int max) |
629 |
|
{ |
630 |
|
int q,i; |
631 |
|
int64_t timeSTART; |
632 |
|
int64_t timeA = 0; |
633 |
|
int64_t timeB = 0; |
634 |
|
DECLARE_ALIGNED_MATRIX(arrayX, 1, 64, int16_t, CACHE_LINE); |
635 |
|
DECLARE_ALIGNED_MATRIX(arrayA, 1, 64, int16_t, CACHE_LINE); |
636 |
|
DECLARE_ALIGNED_MATRIX(arrayB, 1, 64, int16_t, CACHE_LINE); |
637 |
|
|
638 |
|
for (q = 1; q <= 31; q++) /* quantizer */ |
639 |
|
{ |
640 |
|
for (i = min; i < max; i++) /* input coeff */ |
641 |
|
{ |
642 |
|
fill16(arrayX, 64, i); |
643 |
|
|
644 |
|
timeSTART = read_counter(); |
645 |
|
funcA(arrayA, arrayX, q, q); |
646 |
|
timeA += read_counter() - timeSTART; |
647 |
|
|
648 |
|
timeSTART = read_counter(); |
649 |
|
funcB(arrayB, arrayX, q, q); |
650 |
|
timeB += read_counter() - timeSTART; |
651 |
|
|
652 |
|
if (compare16(arrayA, arrayB, 64)) |
653 |
|
{ |
654 |
|
printf("%s/%s error: q=%i, i=%i\n", nameA?nameA:"?", nameB?nameB:"?", q, i); |
655 |
|
return 0; |
656 |
|
} |
657 |
|
} |
658 |
|
} |
659 |
|
|
660 |
|
if (nameA) printf("%s:\t%I64i\n", nameA, timeA); |
661 |
|
if (nameB) printf("%s:\t%I64i\n", nameB, timeB); |
662 |
|
|
663 |
|
return 0; |
664 |
|
} |
665 |
|
|
666 |
|
int test_h263_inter(quanth263_interFunc * funcA, quanth263_interFunc * funcB, |
667 |
|
const char * nameA, const char * nameB, |
668 |
|
int min, int max) |
669 |
|
{ |
670 |
|
int q,i; |
671 |
|
int64_t timeSTART; |
672 |
|
int64_t timeA = 0; |
673 |
|
int64_t timeB = 0; |
674 |
|
DECLARE_ALIGNED_MATRIX(arrayX, 1, 64, int16_t, CACHE_LINE); |
675 |
|
DECLARE_ALIGNED_MATRIX(arrayA, 1, 64, int16_t, CACHE_LINE); |
676 |
|
DECLARE_ALIGNED_MATRIX(arrayB, 1, 64, int16_t, CACHE_LINE); |
677 |
|
|
678 |
|
for (q = 1; q <= 31; q++) /* quantizer */ |
679 |
|
{ |
680 |
|
for (i = min; i < max; i++) /* input coeff */ |
681 |
|
{ |
682 |
|
fill16(arrayX, 64, i); |
683 |
|
|
684 |
|
timeSTART = read_counter(); |
685 |
|
funcA(arrayA, arrayX, q); |
686 |
|
timeA += read_counter() - timeSTART; |
687 |
|
|
688 |
|
timeSTART = read_counter(); |
689 |
|
funcB(arrayB, arrayX, q); |
690 |
|
timeB += read_counter() - timeSTART; |
691 |
|
|
692 |
|
if (compare16(arrayA, arrayB, 64)) |
693 |
|
{ |
694 |
|
printf("%s/%s error: q=%i, i=%i\n", nameA?nameA:"?", nameB?nameB:"?", q, i); |
695 |
|
return 0; |
696 |
|
} |
697 |
|
} |
698 |
|
} |
699 |
|
|
700 |
|
if (nameA) printf("%s:\t%I64i\n", nameA, timeA); |
701 |
|
if (nameB) printf("%s:\t%I64i\n", nameB, timeB); |
702 |
|
|
703 |
|
return 0; |
704 |
|
} |
705 |
|
|
706 |
|
|
707 |
|
|
708 |
|
int xvid_init_test() |
709 |
|
{ |
710 |
|
int cpu_flags; |
711 |
|
|
712 |
|
printf("xvid_init_test\n"); |
713 |
|
|
714 |
|
#if defined(ARCH_X86) |
715 |
|
cpu_flags = check_cpu_features(); |
716 |
|
|
717 |
|
emms_mmx(); |
718 |
|
|
719 |
|
printf("--- quant intra ---\n"); |
720 |
|
if (cpu_flags & XVID_CPU_MMX) |
721 |
|
test_h263_intra(quant_intra_c, quant_intra_mmx, "c", "mmx", -2048, 2047); |
722 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
723 |
|
test_h263_intra(quant_intra_c, quant_intra_3dne, NULL, "3dne", -2048, 2047); |
724 |
|
if (cpu_flags & XVID_CPU_SSE2) |
725 |
|
test_h263_intra(quant_intra_c, quant_intra_sse2, NULL, "sse2", -2048, 2047); |
726 |
|
|
727 |
|
printf("\n--- quant inter ---\n"); |
728 |
|
if (cpu_flags & XVID_CPU_MMX) |
729 |
|
test_h263_inter(quant_inter_c, quant_inter_mmx, "c", "mmx", -2048, 2047); |
730 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
731 |
|
test_h263_inter(quant_inter_c, quant_inter_3dne, NULL, "3dne", -2048, 2047); |
732 |
|
if (cpu_flags & XVID_CPU_SSE2) |
733 |
|
test_h263_inter(quant_inter_c, quant_inter_sse2, NULL, "sse2", -2048, 2047); |
734 |
|
|
735 |
|
printf("\n--- dequan intra ---\n"); |
736 |
|
if (cpu_flags & XVID_CPU_MMX) |
737 |
|
test_h263_intra(dequant_intra_c, dequant_intra_mmx, "c", "mmx", -256, 255); |
738 |
|
if (cpu_flags & XVID_CPU_MMXEXT) |
739 |
|
test_h263_intra(dequant_intra_c, dequant_intra_xmm, NULL, "xmm", -256, 255); |
740 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
741 |
|
test_h263_intra(dequant_intra_c, dequant_intra_3dne, NULL, "3dne", -256, 255); |
742 |
|
if (cpu_flags & XVID_CPU_SSE2) |
743 |
|
test_h263_intra(dequant_intra_c, dequant_intra_sse2, NULL, "sse2", -256, 255); |
744 |
|
|
745 |
|
printf("\n--- dequant inter ---\n"); |
746 |
|
if (cpu_flags & XVID_CPU_MMX) |
747 |
|
test_h263_inter((quanth263_interFunc*)dequant_inter_c, |
748 |
|
(quanth263_interFunc*)dequant_inter_mmx, "c", "mmx", -256, 255); |
749 |
|
|
750 |
|
if (cpu_flags & XVID_CPU_MMXEXT) |
751 |
|
test_h263_inter((quanth263_interFunc*)dequant_inter_c, |
752 |
|
(quanth263_interFunc*)dequant_inter_xmm, NULL, "xmm", -256, 255); |
753 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
754 |
|
test_h263_inter((quanth263_interFunc*)dequant_inter_c, |
755 |
|
(quanth263_interFunc*)dequant_inter_3dne, NULL, "3dne", -256, 255); |
756 |
|
if (cpu_flags & XVID_CPU_SSE2) |
757 |
|
test_h263_inter((quanth263_interFunc*)dequant_inter_c, |
758 |
|
(quanth263_interFunc*)dequant_inter_sse2, NULL, "sse2", -256, 255); |
759 |
|
|
760 |
|
printf("\n--- quant4_intra ---\n"); |
761 |
|
if (cpu_flags & XVID_CPU_MMX) |
762 |
|
test_h263_intra((quanth263_intraFunc*)quant4_intra_c, |
763 |
|
(quanth263_intraFunc*)quant4_intra_mmx, "c", "mmx", -2048, 2047); |
764 |
|
if (cpu_flags & XVID_CPU_MMXEXT) |
765 |
|
test_h263_intra((quanth263_intraFunc*)quant4_intra_c, |
766 |
|
(quanth263_intraFunc*)quant4_intra_xmm, NULL, "xmm", -2048, 2047); |
767 |
|
|
768 |
|
printf("\n--- quant4_inter ---\n"); |
769 |
|
if (cpu_flags & XVID_CPU_MMX) |
770 |
|
test_h263_inter((quanth263_interFunc*)quant4_inter_c, |
771 |
|
(quanth263_interFunc*)quant4_inter_mmx, "c", "mmx", -2048, 2047); |
772 |
|
if (cpu_flags & XVID_CPU_MMXEXT) |
773 |
|
test_h263_inter((quanth263_interFunc*)quant4_inter_c, |
774 |
|
(quanth263_interFunc*)quant4_inter_xmm, NULL, "xmm", -2048, 2047); |
775 |
|
|
776 |
|
|
777 |
|
printf("\n--- dequant4_intra ---\n"); |
778 |
|
if (cpu_flags & XVID_CPU_MMX) |
779 |
|
test_h263_intra((quanth263_intraFunc*)dequant4_intra_c, |
780 |
|
(quanth263_intraFunc*)dequant4_intra_mmx, "c", "mmx", -256, 255); |
781 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
782 |
|
test_h263_intra((quanth263_intraFunc*)dequant4_intra_c, |
783 |
|
(quanth263_intraFunc*)dequant4_intra_3dne, NULL, "sse2", -256, 255); |
784 |
|
|
785 |
|
printf("\n--- dequant4_inter ---\n"); |
786 |
|
if (cpu_flags & XVID_CPU_MMX) |
787 |
|
test_h263_inter((quanth263_interFunc*)dequant4_inter_c, |
788 |
|
(quanth263_interFunc*)dequant4_inter_mmx, "c", "mmx", -256, 255); |
789 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
790 |
|
test_h263_inter((quanth263_interFunc*)dequant4_inter_c, |
791 |
|
(quanth263_interFunc*)dequant4_inter_3dne, NULL, "sse2", -256, 255); |
792 |
|
|
793 |
|
emms_mmx(); |
794 |
|
|
795 |
|
#endif |
796 |
|
|
797 |
|
return XVID_ERR_OK; |
798 |
|
} |
799 |
|
|
800 |
|
|
801 |
int |
int |
802 |
xvid_init(void *handle, |
xvid_init(void *handle, |
803 |
int opt, |
int opt, |
812 |
case XVID_INIT_CONVERT : |
case XVID_INIT_CONVERT : |
813 |
return xvid_init_convert((XVID_INIT_CONVERTINFO*)param1); |
return xvid_init_convert((XVID_INIT_CONVERTINFO*)param1); |
814 |
|
|
815 |
|
case XVID_INIT_TEST : |
816 |
|
return xvid_init_test(); |
817 |
|
|
818 |
default : |
default : |
819 |
return XVID_ERR_FAIL; |
return XVID_ERR_FAIL; |
820 |
} |
} |