Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(482)

Issues Search
    My Issues | Starred     Open | Closed | All

Side by Side Diff: source/row_mips.cc

Issue 2595333002: Libyuv MIPS DSPR2 optimizations. (Closed)
Patch Set: Created 3 years, 12 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View unified diff | Download patch
« no previous file with comments | « source/row_any.cc ('k') | source/scale.cc » ('j') | no next file with comments »
Toggle Intra-line Diffs ('i') | Expand Comments ('e') | Collapse Comments ('c') | Show Comments Hide Comments ('s')
OLDNEW
1 /* 1 /*
2 * Copyright (c) 2012 The LibYuv project authors. All Rights Reserved. 2 * Copyright (c) 2012 The LibYuv project authors. All Rights Reserved.
3 * 3 *
4 * Use of this source code is governed by a BSD-style license 4 * Use of this source code is governed by a BSD-style license
5 * that can be found in the LICENSE file in the root of the source 5 * that can be found in the LICENSE file in the root of the source
6 * tree. An additional intellectual property rights grant can be found 6 * tree. An additional intellectual property rights grant can be found
7 * in the file PATENTS. All contributing project authors may 7 * in the file PATENTS. All contributing project authors may
8 * be found in the AUTHORS file in the root of the source tree. 8 * be found in the AUTHORS file in the root of the source tree.
9 */ 9 */
10 10
(...skipping 567 matching lines...) Expand 10 before | Expand all | Expand 10 after
578 " addiu %[dst_v], %[dst_v], 1 \n" 578 " addiu %[dst_v], %[dst_v], 1 \n"
579 579
580 "3: \n" 580 "3: \n"
581 ".set pop \n" 581 ".set pop \n"
582 : [src_uv] "+r"(src_uv), [dst_u] "+r"(dst_u), [dst_v] "+r"(dst_v), 582 : [src_uv] "+r"(src_uv), [dst_u] "+r"(dst_u), [dst_v] "+r"(dst_v),
583 [x] "=&r"(x), [y] "=&r"(y) 583 [x] "=&r"(x), [y] "=&r"(y)
584 : [width] "r"(width) 584 : [width] "r"(width)
585 : "t0", "t1", "t2", "t3", "t4", "t5", "t7", "t8", "t9"); 585 : "t0", "t1", "t2", "t3", "t4", "t5", "t7", "t8", "t9");
586 } 586 }
587 587
588 // Convert (4 Y and 2 VU) I422 and arrange RGB values into 588 void I422ToARGBRow_DSPR2(const uint8* src_y,
589 // t5 = | 0 | B0 | 0 | b0 | 589 const uint8* src_u,
590 // t4 = | 0 | B1 | 0 | b1 | 590 const uint8* src_v,
591 // t9 = | 0 | G0 | 0 | g0 | 591 uint8* rgb_buf,
592 // t8 = | 0 | G1 | 0 | g1 | 592 const struct YuvConstants* yuvconstants,
593 // t2 = | 0 | R0 | 0 | r0 | 593 int width) {
594 // t1 = | 0 | R1 | 0 | r1 | 594 int x;
595 #define YUVTORGB \ 595 uint32 tmp_ub = yuvconstants->kUVToB[0];
596 "lw $t0, 0(%[y_buf]) \n" \ 596 uint32 tmp_ug = yuvconstants->kUVToG[0];
597 "lhu $t1, 0(%[u_buf]) \n" \ 597 uint32 tmp_vg = yuvconstants->kUVToG[1];
598 "lhu $t2, 0(%[v_buf]) \n" \ 598 uint32 tmp_vr = yuvconstants->kUVToR[1];
599 "preceu.ph.qbr $t1, $t1 \n" \ 599 uint32 tmp_bb = yuvconstants->kUVBiasB[0];
600 "preceu.ph.qbr $t2, $t2 \n" \ 600 uint32 tmp_bg = yuvconstants->kUVBiasG[0];
601 "preceu.ph.qbra $t3, $t0 \n" \ 601 uint32 tmp_br = yuvconstants->kUVBiasR[0];
602 "preceu.ph.qbla $t0, $t0 \n" \ 602 uint32 yg = yuvconstants->kYToRgb[0];
603 "subu.ph $t1, $t1, $s5 \n" \ 603 uint32 tmp_yg;
604 "subu.ph $t2, $t2, $s5 \n" \ 604 uint32 tmp_mask = 0x7fff7fff;
605 "subu.ph $t3, $t3, $s4 \n" \ 605 tmp_bb = ((uint)(tmp_bb & 0xffff) << 16) | (tmp_bb & 0xffff);
606 "subu.ph $t0, $t0, $s4 \n" \ 606 tmp_bg = ((uint)(tmp_bg & 0xffff) << 16) | (tmp_bg & 0xffff);
607 "mul.ph $t3, $t3, $s0 \n" \ 607 tmp_br = ((uint)(tmp_br & 0xffff) << 16) | (tmp_br & 0xffff);
608 "mul.ph $t0, $t0, $s0 \n" \ 608 tmp_yg = ((uint)(yg & 0xffff) << 16) | (yg & 0xffff);
609 "shll.ph $t4, $t1, 0x7 \n" \ 609 tmp_ub = ~(((uint)(tmp_ub & 0xffff) << 16) | (tmp_ub & 0xffff)) + 0x00010001;
610 "subu.ph $t4, $t4, $t1 \n" \ 610 tmp_ug = ((uint)(tmp_ug & 0xffff) << 16) | (tmp_ug & 0xffff);
611 "mul.ph $t6, $t1, $s1 \n" \ 611 tmp_vg = ((uint)(tmp_vg & 0xffff) << 16) | (tmp_vg & 0xffff);
612 "mul.ph $t1, $t2, $s2 \n" \ 612 tmp_vr = ~(((uint)(tmp_vr & 0xffff) << 16) | (tmp_vr & 0xffff)) + 0x00010001;
613 "addq_s.ph $t5, $t4, $t3 \n" \ 613 yg = yg * 0x0101;
614 "addq_s.ph $t4, $t4, $t0 \n" \ 614
615 "shra.ph $t5, $t5, 6 \n" \ 615 for (x = 0; x < width - 1; x += 2) {
616 "shra.ph $t4, $t4, 6 \n" \ 616 uint32 tmp_t1, tmp_t2, tmp_t3, tmp_t4, tmp_t5;
617 "addiu %[u_buf], 2 \n" \ 617 uint32 tmp_t6, tmp_t7, tmp_t8, tmp_t9;
618 "addiu %[v_buf], 2 \n" \ 618 __asm__ __volatile__ (
619 "addu.ph $t6, $t6, $t1 \n" \ 619 ".set push \n"
620 "mul.ph $t1, $t2, $s3 \n" \ 620 ".set noreorder \n"
621 "addu.ph $t9, $t6, $t3 \n" \ 621 "lbu %[tmp_t7], 0(%[src_y]) \n"
622 "addu.ph $t8, $t6, $t0 \n" \ 622 "lbu %[tmp_t1], 1(%[src_y]) \n"
623 "shra.ph $t9, $t9, 6 \n" \ 623 "mul %[tmp_t7], %[tmp_t7], %[yg] \n"
624 "shra.ph $t8, $t8, 6 \n" \ 624 "mul %[tmp_t1], %[tmp_t1], %[yg] \n"
625 "addu.ph $t2, $t1, $t3 \n" \ 625 "lbu %[tmp_t2], 0(%[src_u]) \n"
626 "addu.ph $t1, $t1, $t0 \n" \ 626 "lbu %[tmp_t3], 0(%[src_v]) \n"
627 "shra.ph $t2, $t2, 6 \n" \ 627 "replv.ph %[tmp_t2], %[tmp_t2] \n"
628 "shra.ph $t1, $t1, 6 \n" \ 628 "replv.ph %[tmp_t3], %[tmp_t3] \n"
629 "subu.ph $t5, $t5, $s5 \n" \ 629 "mul.ph %[tmp_t4], %[tmp_t2], %[tmp_ub] \n"
630 "subu.ph $t4, $t4, $s5 \n" \ 630 "mul.ph %[tmp_t5], %[tmp_t2], %[tmp_ug] \n"
631 "subu.ph $t9, $t9, $s5 \n" \ 631 "mul.ph %[tmp_t6], %[tmp_t3], %[tmp_vr] \n"
632 "subu.ph $t8, $t8, $s5 \n" \ 632 "mul.ph %[tmp_t3], %[tmp_t3], %[tmp_vg] \n"
633 "subu.ph $t2, $t2, $s5 \n" \ 633 "srl %[tmp_t7], %[tmp_t7], 16 \n"
634 "subu.ph $t1, $t1, $s5 \n" \ 634 "ins %[tmp_t1], %[tmp_t7], 0, 16 \n"
635 "shll_s.ph $t5, $t5, 8 \n" \ 635 "addq_s.ph %[tmp_t7], %[tmp_t1], %[tmp_bb] \n"
636 "shll_s.ph $t4, $t4, 8 \n" \ 636 "addq_s.ph %[tmp_t8], %[tmp_t1], %[tmp_bg] \n"
637 "shll_s.ph $t9, $t9, 8 \n" \ 637 "addq_s.ph %[tmp_t9], %[tmp_t1], %[tmp_br] \n"
638 "shll_s.ph $t8, $t8, 8 \n" \ 638 "addq_s.ph %[tmp_t5], %[tmp_t5], %[tmp_t3] \n"
639 "shll_s.ph $t2, $t2, 8 \n" \ 639 "addq_s.ph %[tmp_t7], %[tmp_t7], %[tmp_t4] \n"
640 "shll_s.ph $t1, $t1, 8 \n" \ 640 "subq_s.ph %[tmp_t8], %[tmp_t8], %[tmp_t5] \n"
641 "shra.ph $t5, $t5, 8 \n" \ 641 "addq_s.ph %[tmp_t9], %[tmp_t9], %[tmp_t6] \n"
642 "shra.ph $t4, $t4, 8 \n" \ 642 "shra.ph %[tmp_t7], %[tmp_t7], 6 \n"
643 "shra.ph $t9, $t9, 8 \n" \ 643 "shra.ph %[tmp_t8], %[tmp_t8], 6 \n"
644 "shra.ph $t8, $t8, 8 \n" \ 644 "shra.ph %[tmp_t9], %[tmp_t9], 6 \n"
645 "shra.ph $t2, $t2, 8 \n" \ 645 "shll_s.ph %[tmp_t7], %[tmp_t7], 7 \n"
646 "shra.ph $t1, $t1, 8 \n" \ 646 "shll_s.ph %[tmp_t8], %[tmp_t8], 7 \n"
647 "addu.ph $t5, $t5, $s5 \n" \ 647 "shll_s.ph %[tmp_t9], %[tmp_t9], 7 \n"
648 "addu.ph $t4, $t4, $s5 \n" \ 648 "precrqu_s.qb.ph %[tmp_t8], %[tmp_mask], %[tmp_t8] \n"
649 "addu.ph $t9, $t9, $s5 \n" \ 649 "precrqu_s.qb.ph %[tmp_t7], %[tmp_t9], %[tmp_t7] \n"
650 "addu.ph $t8, $t8, $s5 \n" \ 650 "precrq.ph.w %[tmp_t9], %[tmp_t8], %[tmp_t7] \n"
651 "addu.ph $t2, $t2, $s5 \n" \ 651 "ins %[tmp_t7], %[tmp_t8], 16, 16 \n"
652 "addu.ph $t1, $t1, $s5 \n" 652 "precr.qb.ph %[tmp_t8], %[tmp_t9], %[tmp_t7] \n"
653 653 "precrq.qb.ph %[tmp_t7], %[tmp_t9], %[tmp_t7] \n"
654 // TODO(fbarchard): accept yuv conversion constants. 654 "sw %[tmp_t8], 0(%[rgb_buf]) \n"
655 void I422ToARGBRow_DSPR2(const uint8* y_buf, 655 "sw %[tmp_t7], 4(%[rgb_buf]) \n"
656 ".set pop \n"
657 :[tmp_t1] "=&r" (tmp_t1), [tmp_t2] "=&r" (tmp_t2),
658 [tmp_t3] "=&r" (tmp_t3), [tmp_t4] "=&r" (tmp_t4),
659 [tmp_t5] "=&r" (tmp_t5), [tmp_t6] "=&r" (tmp_t6),
660 [tmp_t7] "=&r" (tmp_t7), [tmp_t8] "=&r" (tmp_t8),
661 [tmp_t9] "=&r" (tmp_t9)
662 :[src_y] "r" (src_y), [src_u] "r" (src_u),[src_v] "r" (src_v),
663 [tmp_ub] "r" (tmp_ub), [tmp_ug] "r" (tmp_ug), [yg] "r" (yg),
664 [tmp_vg] "r" (tmp_vg), [tmp_vr] "r" (tmp_vr),
665 [tmp_bb] "r" (tmp_bb), [tmp_bg] "r" (tmp_bg),
666 [tmp_br] "r" (tmp_br), [tmp_yg] "r" (tmp_yg),
667 [rgb_buf] "r" (rgb_buf), [tmp_mask] "r" (tmp_mask)
668 );
669 src_y += 2;
670 src_u += 1;
671 src_v += 1;
672 rgb_buf += 8; // Advance 4 pixels.
673 }
674 }
675
676 // Bilinear filter 8x2 -> 8x1
677 void InterpolateRow_DSPR2(uint8* dst_ptr, const uint8* src_ptr,
678 ptrdiff_t src_stride, int dst_width,
679 int source_y_fraction) {
680 int y0_fraction = 256 - source_y_fraction;
681 const uint8* src_ptr1 = src_ptr + src_stride;
682
683 __asm__ __volatile__ (
684 ".set push \n"
685 ".set noreorder \n"
686
687 "replv.ph $t0, %[y0_fraction] \n"
688 "replv.ph $t1, %[source_y_fraction] \n"
689
690 "1: \n"
691 "lw $t2, 0(%[src_ptr]) \n"
692 "lw $t3, 0(%[src_ptr1]) \n"
693 "lw $t4, 4(%[src_ptr]) \n"
694 "lw $t5, 4(%[src_ptr1]) \n"
695 "muleu_s.ph.qbl $t6, $t2, $t0 \n"
696 "muleu_s.ph.qbr $t7, $t2, $t0 \n"
697 "muleu_s.ph.qbl $t8, $t3, $t1 \n"
698 "muleu_s.ph.qbr $t9, $t3, $t1 \n"
699 "muleu_s.ph.qbl $t2, $t4, $t0 \n"
700 "muleu_s.ph.qbr $t3, $t4, $t0 \n"
701 "muleu_s.ph.qbl $t4, $t5, $t1 \n"
702 "muleu_s.ph.qbr $t5, $t5, $t1 \n"
703 "addq.ph $t6, $t6, $t8 \n"
704 "addq.ph $t7, $t7, $t9 \n"
705 "addq.ph $t2, $t2, $t4 \n"
706 "addq.ph $t3, $t3, $t5 \n"
707 "shra_r.ph $t6, $t6, 8 \n"
708 "shra_r.ph $t7, $t7, 8 \n"
709 "shra_r.ph $t2, $t2, 8 \n"
710 "shra_r.ph $t3, $t3, 8 \n"
711 "precr.qb.ph $t6, $t6, $t7 \n"
712 "precr.qb.ph $t2, $t2, $t3 \n"
713 "addiu %[src_ptr], %[src_ptr], 8 \n"
714 "addiu %[src_ptr1], %[src_ptr1], 8 \n"
715 "addiu %[dst_width], %[dst_width], -8 \n"
716 "sw $t6, 0(%[dst_ptr]) \n"
717 "sw $t2, 4(%[dst_ptr]) \n"
718 "bgtz %[dst_width], 1b \n"
719 " addiu %[dst_ptr], %[dst_ptr], 8 \n"
720
721 ".set pop \n"
722 : [dst_ptr] "+r" (dst_ptr),
723 [src_ptr1] "+r" (src_ptr1),
724 [src_ptr] "+r" (src_ptr),
725 [dst_width] "+r" (dst_width)
726 : [source_y_fraction] "r" (source_y_fraction),
727 [y0_fraction] "r" (y0_fraction),
728 [src_stride] "r" (src_stride)
729 : "t0", "t1", "t2", "t3", "t4", "t5",
730 "t6", "t7", "t8", "t9"
731 );
732 }
733 #include <stdio.h>
734 void RGB24ToARGBRow_DSPR2(const uint8* src_rgb24, uint8* dst_argb, int width) {
735 int x;
736 uint32 tmp_mask = 0xff;
737 uint32 tmp_t1;
738 for (x = 0; x < (width - 1); ++x) {
739 __asm__ __volatile__ (
740 ".set push \n"
741 ".set noreorder \n"
742 "ulw %[tmp_t1], 0(%[src_rgb24]) \n"
743 "addiu %[dst_argb], %[dst_argb], 4 \n"
744 "addiu %[src_rgb24], %[src_rgb24], 3 \n"
745 "ins %[tmp_t1], %[tmp_mask], 24, 8 \n"
746 "sw %[tmp_t1], -4(%[dst_argb]) \n"
747 ".set pop \n"
748 :[src_rgb24] "+r" (src_rgb24), [dst_argb] "+r" (dst_argb),
749 [tmp_t1] "=&r" (tmp_t1)
750 :[tmp_mask] "r" (tmp_mask)
751 : "memory"
752 );
753 }
754 uint8 b = src_rgb24[0];
755 uint8 g = src_rgb24[1];
756 uint8 r = src_rgb24[2];
757 dst_argb[0] = b;
758 dst_argb[1] = g;
759 dst_argb[2] = r;
760 dst_argb[3] = 255u;
761 }
762
763 void RAWToARGBRow_DSPR2(const uint8* src_raw, uint8* dst_argb, int width) {
764 int x;
765 uint32 tmp_mask = 0xff;
766 uint32 tmp_t1, tmp_t2;
767 for (x = 0; x < (width-1); ++x) {
768 __asm__ __volatile__ (
769 ".set push \n"
770 ".set noreorder \n"
771 "ulw %[tmp_t1], 0(%[src_raw]) \n"
772 "addiu %[dst_argb], %[dst_argb], 4 \n"
773 "addiu %[src_raw], %[src_raw], 3 \n"
774 "srl %[tmp_t2], %[tmp_t1], 16 \n"
775 "ins %[tmp_t1], %[tmp_mask], 24, 8 \n"
776 "ins %[tmp_t1], %[tmp_t1], 16, 8 \n"
777 "ins %[tmp_t1], %[tmp_t2], 0, 8 \n"
778 "sw %[tmp_t1], -4(%[dst_argb]) \n"
779 ".set pop \n"
780 :[src_raw] "+r" (src_raw), [dst_argb] "+r" (dst_argb),
781 [tmp_t1] "=&r" (tmp_t1), [tmp_t2] "=&r" (tmp_t2)
782 :[tmp_mask] "r" (tmp_mask)
783 : "memory"
784 );
785 }
786 uint8 r = src_raw[0];
787 uint8 g = src_raw[1];
788 uint8 b = src_raw[2];
789 dst_argb[0] = b;
790 dst_argb[1] = g;
791 dst_argb[2] = r;
792 dst_argb[3] = 255u;
793 }
794
795 void RGB565ToARGBRow_DSPR2(const uint8* src_rgb565, uint8* dst_argb,
796 int width) {
797 int x;
798 uint32 tmp_mask = 0xff;
799 uint32 tmp_t1, tmp_t2, tmp_t3;
800 for (x = 0; x < width; ++x) {
801 __asm__ __volatile__ (
802 ".set push \n"
803 ".set noreorder \n"
804 "lhu %[tmp_t1], 0(%[src_rgb565]) \n"
805 "addiu %[dst_argb], %[dst_argb], 4 \n"
806 "addiu %[src_rgb565], %[src_rgb565], 2 \n"
807 "sll %[tmp_t2], %[tmp_t1], 8 \n"
808 "ins %[tmp_t2], %[tmp_mask], 24,8 \n"
809 "ins %[tmp_t2], %[tmp_t1], 3, 16 \n"
810 "ins %[tmp_t2], %[tmp_t1], 5, 11 \n"
811 "srl %[tmp_t3], %[tmp_t1], 9 \n"
812 "ins %[tmp_t2], %[tmp_t3], 8, 2 \n"
813 "ins %[tmp_t2], %[tmp_t1], 3, 5 \n"
814 "srl %[tmp_t3], %[tmp_t1], 2 \n"
815 "ins %[tmp_t2], %[tmp_t3], 0, 3 \n"
816 "sw %[tmp_t2], -4(%[dst_argb]) \n"
817 ".set pop \n"
818 :[tmp_t1] "=&r" (tmp_t1), [tmp_t2] "=&r" (tmp_t2),
819 [tmp_t3] "=&r" (tmp_t3), [src_rgb565] "+r" (src_rgb565),
820 [dst_argb] "+r" (dst_argb)
821 :[tmp_mask] "r" (tmp_mask)
822 );
823 }
824 }
825
826 void ARGB1555ToARGBRow_DSPR2(const uint8* src_argb1555, uint8* dst_argb,
827 int width) {
828 int x;
829 uint32 tmp_t1, tmp_t2, tmp_t3;
830 for (x = 0; x < width; ++x) {
831 __asm__ __volatile__ (
832 ".set push \n"
833 ".set noreorder \n"
834 "lh %[tmp_t1], 0(%[src_argb1555]) \n"
835 "addiu %[dst_argb], %[dst_argb], 4 \n"
836 "addiu %[src_argb1555], %[src_argb1555], 2 \n"
837 "sll %[tmp_t2], %[tmp_t1], 9 \n"
838 "ins %[tmp_t2], %[tmp_t1], 4, 15 \n"
839 "ins %[tmp_t2], %[tmp_t1], 6, 10 \n"
840 "srl %[tmp_t3], %[tmp_t1], 7 \n"
841 "ins %[tmp_t2], %[tmp_t3], 8, 3 \n"
842 "ins %[tmp_t2], %[tmp_t1], 3, 5 \n"
843 "srl %[tmp_t3], %[tmp_t1], 2 \n"
844 "ins %[tmp_t2], %[tmp_t3], 0, 3 \n"
845 "sw %[tmp_t2], -4(%[dst_argb]) \n"
846 ".set pop \n"
847 :[tmp_t1] "=&r" (tmp_t1), [tmp_t2] "=&r" (tmp_t2),
848 [tmp_t3] "=&r" (tmp_t3), [src_argb1555] "+r" (src_argb1555),
849 [dst_argb] "+r" (dst_argb)
850 :
851 );
852 }
853 }
854
855 void ARGB4444ToARGBRow_DSPR2(const uint8* src_argb4444, uint8* dst_argb,
856 int width) {
857 int x;
858 uint32 tmp_t1;
859 for (x = 0; x < width; ++x) {
860 __asm__ __volatile__ (
861 ".set push \n"
862 ".set noreorder \n"
863 "lh %[tmp_t1], 0(%[src_argb4444]) \n"
864 "addiu %[dst_argb], %[dst_argb], 4 \n"
865 "addiu %[src_argb4444], %[src_argb4444], 2 \n"
866 "ins %[tmp_t1], %[tmp_t1], 16, 16 \n"
867 "ins %[tmp_t1], %[tmp_t1], 12, 16 \n"
868 "ins %[tmp_t1], %[tmp_t1], 8, 12 \n"
869 "ins %[tmp_t1], %[tmp_t1], 4, 8 \n"
870 "sw %[tmp_t1], -4(%[dst_argb]) \n"
871 ".set pop \n"
872 :[src_argb4444] "+r" (src_argb4444), [dst_argb] "+r" (dst_argb),
873 [tmp_t1] "=&r" (tmp_t1)
874 );
875 }
876 }
877
878 void I444ToARGBRow_DSPR2(const uint8* y_buf,
656 const uint8* u_buf, 879 const uint8* u_buf,
657 const uint8* v_buf, 880 const uint8* v_buf,
658 uint8* rgb_buf, 881 uint8* rgb_buf,
659 const struct YuvConstants* yuvconstants, 882 const struct YuvConstants* yuvconstants,
660 int width) { 883 int width) {
661 __asm__ __volatile__( 884 int x;
662 ".set push \n" 885 uint32 tmp_ub = yuvconstants->kUVToB[0];
663 ".set noreorder \n" 886 uint32 tmp_ug = yuvconstants->kUVToG[0];
664 "beqz %[width], 2f \n" 887 uint32 tmp_vg = yuvconstants->kUVToG[1];
665 " repl.ph $s0, 74 \n" // |YG|YG| = |74|74| 888 uint32 tmp_vr = yuvconstants->kUVToR[1];
666 "repl.ph $s1, -25 \n" // |UG|UG| = |-25|-25| 889 uint32 tmp_bb = yuvconstants->kUVBiasB[0];
667 "repl.ph $s2, -52 \n" // |VG|VG| = |-52|-52| 890 uint32 tmp_bg = yuvconstants->kUVBiasG[0];
668 "repl.ph $s3, 102 \n" // |VR|VR| = |102|102| 891 uint32 tmp_br = yuvconstants->kUVBiasR[0];
669 "repl.ph $s4, 16 \n" // |0|16|0|16| 892 uint32 yg = yuvconstants->kYToRgb[0];
670 "repl.ph $s5, 128 \n" // |128|128| // clipping 893 uint32 tmp_mask = 0x7fff7fff;
671 "lui $s6, 0xff00 \n" 894 uint32 tmp_yg;
672 "ori $s6, 0xff00 \n" // |ff|00|ff|00|ff| 895
673 896 tmp_bb = ((uint)(tmp_bb & 0xffff) << 16) | (tmp_bb & 0xffff);
674 "1: \n" YUVTORGB 897 tmp_bg = ((uint)(tmp_bg & 0xffff) << 16) | (tmp_bg & 0xffff);
675 // Arranging into argb format 898 tmp_br = ((uint)(tmp_br & 0xffff) << 16) | (tmp_br & 0xffff);
676 "precr.qb.ph $t4, $t8, $t4 \n" // |G1|g1|B1|b1| 899 tmp_yg = ((uint)(yg & 0xffff) << 16) | (yg & 0xffff);
677 "precr.qb.ph $t5, $t9, $t5 \n" // |G0|g0|B0|b0| 900 tmp_ub = ~(((uint)(tmp_ub & 0xffff) << 16) | (tmp_ub & 0xffff)) + 0x00010001;
678 "addiu %[width], -4 \n" 901 tmp_ug = ((uint)(tmp_ug & 0xffff) << 16) | (tmp_ug & 0xffff);
679 "precrq.qb.ph $t8, $t4, $t5 \n" // |G1|B1|G0|B0| 902 tmp_vg = ((uint)(tmp_vg & 0xffff) << 16) | (tmp_vg & 0xffff);
680 "precr.qb.ph $t9, $t4, $t5 \n" // |g1|b1|g0|b0| 903 tmp_vr = ~(((uint)(tmp_vr & 0xffff) << 16) | (tmp_vr & 0xffff)) + 0x00010001;
681 "precr.qb.ph $t2, $t1, $t2 \n" // |R1|r1|R0|r0| 904 yg = yg * 0x0101;
682 905
683 "addiu %[y_buf], 4 \n" 906 for (x = 0; x < width-1; x+=2) {
684 "preceu.ph.qbla $t1, $t2 \n" // |0 |R1|0 |R0| 907 uint32 tmp_t1, tmp_t2, tmp_t3, tmp_t4, tmp_t5;
685 "preceu.ph.qbra $t2, $t2 \n" // |0 |r1|0 |r0| 908 uint32 tmp_t6, tmp_t7, tmp_t8, tmp_t9;
686 "or $t1, $t1, $s6 \n" // |ff|R1|ff|R0| 909 __asm__ __volatile__ (
687 "or $t2, $t2, $s6 \n" // |ff|r1|ff|r0| 910 ".set push \n"
688 "precrq.ph.w $t0, $t2, $t9 \n" // |ff|r1|g1|b1| 911 ".set noreorder \n"
689 "precrq.ph.w $t3, $t1, $t8 \n" // |ff|R1|G1|B1| 912 "lbu %[tmp_t7], 0(%[y_buf]) \n"
690 "sll $t9, $t9, 16 \n" 913 "lbu %[tmp_t1], 1(%[y_buf]) \n"
691 "sll $t8, $t8, 16 \n" 914 "mul %[tmp_t7], %[tmp_t7], %[yg] \n"
692 "packrl.ph $t2, $t2, $t9 \n" // |ff|r0|g0|b0| 915 "mul %[tmp_t1], %[tmp_t1], %[yg] \n"
693 "packrl.ph $t1, $t1, $t8 \n" // |ff|R0|G0|B0| 916 "lh %[tmp_t2], 0(%[u_buf]) \n"
694 // Store results. 917 "lh %[tmp_t3], 0(%[v_buf]) \n"
695 "sw $t2, 0(%[rgb_buf]) \n" 918 "preceu.ph.qbr %[tmp_t2], %[tmp_t2] \n"
696 "sw $t0, 4(%[rgb_buf]) \n" 919 "preceu.ph.qbr %[tmp_t3], %[tmp_t3] \n"
697 "sw $t1, 8(%[rgb_buf]) \n" 920 "mul.ph %[tmp_t4], %[tmp_t2], %[tmp_ub] \n"
698 "sw $t3, 12(%[rgb_buf]) \n" 921 "mul.ph %[tmp_t5], %[tmp_t2], %[tmp_ug] \n"
699 "bnez %[width], 1b \n" 922 "mul.ph %[tmp_t6], %[tmp_t3], %[tmp_vr] \n"
700 " addiu %[rgb_buf], 16 \n" 923 "mul.ph %[tmp_t3], %[tmp_t3], %[tmp_vg] \n"
701 "2: \n" 924 "srl %[tmp_t7], %[tmp_t7], 16 \n"
702 ".set pop \n" 925 "ins %[tmp_t1], %[tmp_t7], 0, 16 \n"
703 : [y_buf] "+r"(y_buf), [u_buf] "+r"(u_buf), [v_buf] "+r"(v_buf), 926 "addq_s.ph %[tmp_t7], %[tmp_t1], %[tmp_bb] \n"
704 [width] "+r"(width), [rgb_buf] "+r"(rgb_buf) 927 "addq_s.ph %[tmp_t8], %[tmp_t1], %[tmp_bg] \n"
705 : 928 "addq_s.ph %[tmp_t9], %[tmp_t1], %[tmp_br] \n"
706 : "t0", "t1", "t2", "t3", "t4", "t5", "t6", "t7", "t8", "t9", "s0", "s1", 929 "addq_s.ph %[tmp_t5], %[tmp_t5], %[tmp_t3] \n"
707 "s2", "s3", "s4", "s5", "s6"); 930 "addq_s.ph %[tmp_t7], %[tmp_t7], %[tmp_t4] \n"
708 } 931 "subq_s.ph %[tmp_t8], %[tmp_t8], %[tmp_t5] \n"
709 932 "addq_s.ph %[tmp_t9], %[tmp_t9], %[tmp_t6] \n"
710 // Bilinear filter 8x2 -> 8x1 933 "shra.ph %[tmp_t7], %[tmp_t7], 6 \n"
711 void InterpolateRow_DSPR2(uint8* dst_ptr, 934 "shra.ph %[tmp_t8], %[tmp_t8], 6 \n"
712 const uint8* src_ptr, 935 "shra.ph %[tmp_t9], %[tmp_t9], 6 \n"
713 ptrdiff_t src_stride, 936 "shll_s.ph %[tmp_t7], %[tmp_t7], 7 \n"
714 int dst_width, 937 "shll_s.ph %[tmp_t8], %[tmp_t8], 7 \n"
715 int source_y_fraction) { 938 "shll_s.ph %[tmp_t9], %[tmp_t9], 7 \n"
716 int y0_fraction = 256 - source_y_fraction; 939 "precrqu_s.qb.ph %[tmp_t8], %[tmp_mask], %[tmp_t8] \n"
717 const uint8* src_ptr1 = src_ptr + src_stride; 940 "precrqu_s.qb.ph %[tmp_t7], %[tmp_t9], %[tmp_t7] \n"
718 941 "precrq.ph.w %[tmp_t2], %[tmp_t8], %[tmp_t7] \n"
719 __asm__ __volatile__( 942 "ins %[tmp_t7], %[tmp_t8], 16, 16 \n"
720 ".set push \n" 943 "precr.qb.ph %[tmp_t8], %[tmp_t2], %[tmp_t7] \n"
721 ".set noreorder \n" 944 "precrq.qb.ph %[tmp_t7], %[tmp_t2], %[tmp_t7] \n"
722 945 "sw %[tmp_t8], 0(%[rgb_buf]) \n"
723 "replv.ph $t0, %[y0_fraction] \n" 946 "sw %[tmp_t7], 4(%[rgb_buf]) \n"
724 "replv.ph $t1, %[source_y_fraction] \n" 947 ".set pop \n"
725 948 :[tmp_t1] "=&r" (tmp_t1), [tmp_t2] "=&r" (tmp_t2),
726 "1: \n" 949 [tmp_t3] "=&r" (tmp_t3), [tmp_t4] "=&r" (tmp_t4),
727 "lw $t2, 0(%[src_ptr]) \n" 950 [tmp_t5] "=&r" (tmp_t5), [tmp_t6] "=&r" (tmp_t6),
728 "lw $t3, 0(%[src_ptr1]) \n" 951 [tmp_t7] "=&r" (tmp_t7), [tmp_t8] "=&r" (tmp_t8),
729 "lw $t4, 4(%[src_ptr]) \n" 952 [tmp_t9] "=&r" (tmp_t9)
730 "lw $t5, 4(%[src_ptr1]) \n" 953 :[y_buf] "r" (y_buf), [yg] "r" (yg),
731 "muleu_s.ph.qbl $t6, $t2, $t0 \n" 954 [u_buf] "r" (u_buf),[v_buf] "r" (v_buf),
732 "muleu_s.ph.qbr $t7, $t2, $t0 \n" 955 [tmp_ub] "r" (tmp_ub), [tmp_ug] "r" (tmp_ug),
733 "muleu_s.ph.qbl $t8, $t3, $t1 \n" 956 [tmp_vg] "r" (tmp_vg), [tmp_vr] "r" (tmp_vr),
734 "muleu_s.ph.qbr $t9, $t3, $t1 \n" 957 [tmp_bb] "r" (tmp_bb), [tmp_bg] "r" (tmp_bg),
735 "muleu_s.ph.qbl $t2, $t4, $t0 \n" 958 [tmp_br] "r" (tmp_br), [tmp_yg] "r" (tmp_yg),
736 "muleu_s.ph.qbr $t3, $t4, $t0 \n" 959 [rgb_buf] "r" (rgb_buf), [tmp_mask] "r" (tmp_mask)
737 "muleu_s.ph.qbl $t4, $t5, $t1 \n" 960 );
738 "muleu_s.ph.qbr $t5, $t5, $t1 \n" 961 y_buf += 2;
739 "addq.ph $t6, $t6, $t8 \n" 962 u_buf += 2;
740 "addq.ph $t7, $t7, $t9 \n" 963 v_buf += 2;
741 "addq.ph $t2, $t2, $t4 \n" 964 rgb_buf += 8; // Advance 1 pixel.
742 "addq.ph $t3, $t3, $t5 \n" 965 }
743 "shra.ph $t6, $t6, 8 \n" 966 }
744 "shra.ph $t7, $t7, 8 \n" 967
745 "shra.ph $t2, $t2, 8 \n" 968 void I422ToARGB4444Row_DSPR2(const uint8* src_y,
746 "shra.ph $t3, $t3, 8 \n" 969 const uint8* src_u,
747 "precr.qb.ph $t6, $t6, $t7 \n" 970 const uint8* src_v,
748 "precr.qb.ph $t2, $t2, $t3 \n" 971 uint8* dst_argb4444,
749 "addiu %[src_ptr], %[src_ptr], 8 \n" 972 const struct YuvConstants* yuvconstants,
750 "addiu %[src_ptr1], %[src_ptr1], 8 \n" 973 int width) {
751 "addiu %[dst_width], %[dst_width], -8 \n" 974 int x;
752 "sw $t6, 0(%[dst_ptr]) \n" 975 uint32 tmp_ub = yuvconstants->kUVToB[0];
753 "sw $t2, 4(%[dst_ptr]) \n" 976 uint32 tmp_ug = yuvconstants->kUVToG[0];
754 "bgtz %[dst_width], 1b \n" 977 uint32 tmp_vg = yuvconstants->kUVToG[1];
755 " addiu %[dst_ptr], %[dst_ptr], 8 \n" 978 uint32 tmp_vr = yuvconstants->kUVToR[1];
756 979 uint32 tmp_bb = yuvconstants->kUVBiasB[0];
757 ".set pop \n" 980 uint32 tmp_bg = yuvconstants->kUVBiasG[0];
758 : [dst_ptr] "+r"(dst_ptr), [src_ptr1] "+r"(src_ptr1), 981 uint32 tmp_br = yuvconstants->kUVBiasR[0];
759 [src_ptr] "+r"(src_ptr), [dst_width] "+r"(dst_width) 982 uint32 yg = yuvconstants->kYToRgb[0];
760 : [source_y_fraction] "r"(source_y_fraction), 983 uint32 tmp_yg;
761 [y0_fraction] "r"(y0_fraction), [src_stride] "r"(src_stride) 984 uint32 tmp_mask = 0x7fff7fff;
762 : "t0", "t1", "t2", "t3", "t4", "t5", "t6", "t7", "t8", "t9"); 985 tmp_bb = ((uint)(tmp_bb & 0xffff) << 16) | (tmp_bb & 0xffff);
763 } 986 tmp_bg = ((uint)(tmp_bg & 0xffff) << 16) | (tmp_bg & 0xffff);
987 tmp_br = ((uint)(tmp_br & 0xffff) << 16) | (tmp_br & 0xffff);
988 tmp_yg = ((uint)(yg & 0xffff) << 16) | (yg & 0xffff);
989 tmp_ub = ~(((uint)(tmp_ub & 0xffff) << 16) | (tmp_ub & 0xffff)) + 0x00010001;
990 tmp_ug = ((uint)(tmp_ug & 0xffff) << 16) | (tmp_ug & 0xffff);
991 tmp_vg = ((uint)(tmp_vg & 0xffff) << 16) | (tmp_vg & 0xffff);
992 tmp_vr = ~(((uint)(tmp_vr & 0xffff) << 16) | (tmp_vr & 0xffff)) + 0x00010001;
993 yg = yg * 0x0101;
994
995 for (x = 0; x < width - 1; x += 2) {
996 uint32 tmp_t1, tmp_t2, tmp_t3, tmp_t4, tmp_t5;
997 uint32 tmp_t6, tmp_t7, tmp_t8, tmp_t9;
998 __asm__ __volatile__ (
999 ".set push \n"
1000 ".set noreorder \n"
1001 "lbu %[tmp_t7], 0(%[src_y]) \n"
1002 "lbu %[tmp_t1], 1(%[src_y]) \n"
1003 "mul %[tmp_t7], %[tmp_t7], %[yg] \n"
1004 "mul %[tmp_t1], %[tmp_t1], %[yg] \n"
1005 "lbu %[tmp_t2], 0(%[src_u]) \n"
1006 "lbu %[tmp_t3], 0(%[src_v]) \n"
1007 "replv.ph %[tmp_t2], %[tmp_t2] \n"
1008 "replv.ph %[tmp_t3], %[tmp_t3] \n"
1009 "mul.ph %[tmp_t4], %[tmp_t2], %[tmp_ub] \n"
1010 "mul.ph %[tmp_t5], %[tmp_t2], %[tmp_ug] \n"
1011 "mul.ph %[tmp_t6], %[tmp_t3], %[tmp_vr] \n"
1012 "mul.ph %[tmp_t3], %[tmp_t3], %[tmp_vg] \n"
1013 "srl %[tmp_t7], %[tmp_t7], 16 \n"
1014 "ins %[tmp_t1], %[tmp_t7], 0, 16 \n"
1015 "addq_s.ph %[tmp_t7], %[tmp_t1], %[tmp_bb] \n"
1016 "addq_s.ph %[tmp_t8], %[tmp_t1], %[tmp_bg] \n"
1017 "addq_s.ph %[tmp_t9], %[tmp_t1], %[tmp_br] \n"
1018 "addq_s.ph %[tmp_t5], %[tmp_t5], %[tmp_t3] \n"
1019 "addq_s.ph %[tmp_t7], %[tmp_t7], %[tmp_t4] \n"
1020 "subq_s.ph %[tmp_t8], %[tmp_t8], %[tmp_t5] \n"
1021 "addq_s.ph %[tmp_t9], %[tmp_t9], %[tmp_t6] \n"
1022 "shra.ph %[tmp_t7], %[tmp_t7], 6 \n"
1023 "shra.ph %[tmp_t8], %[tmp_t8], 6 \n"
1024 "shra.ph %[tmp_t9], %[tmp_t9], 6 \n"
1025 "shll_s.ph %[tmp_t7], %[tmp_t7], 7 \n"
1026 "shll_s.ph %[tmp_t8], %[tmp_t8], 7 \n"
1027 "shll_s.ph %[tmp_t9], %[tmp_t9], 7 \n"
1028 "precrqu_s.qb.ph %[tmp_t8], %[tmp_mask], %[tmp_t8] \n"
1029 "precrqu_s.qb.ph %[tmp_t7], %[tmp_t9], %[tmp_t7] \n"
1030 "precrq.ph.w %[tmp_t2], %[tmp_t8], %[tmp_t7] \n"
1031 "ins %[tmp_t7], %[tmp_t8], 16, 16 \n"
1032 "precr.qb.ph %[tmp_t8], %[tmp_t2], %[tmp_t7] \n"
1033 "precrq.qb.ph %[tmp_t7], %[tmp_t2], %[tmp_t7] \n"
1034 "shrl.qb %[tmp_t1], %[tmp_t8], 4 \n"
1035 "shrl.qb %[tmp_t2], %[tmp_t7], 4 \n"
1036 "shrl.ph %[tmp_t8], %[tmp_t1], 4 \n"
1037 "shrl.ph %[tmp_t7], %[tmp_t2], 4 \n"
1038 "or %[tmp_t8], %[tmp_t8], %[tmp_t1] \n"
1039 "or %[tmp_t7], %[tmp_t7], %[tmp_t2] \n"
1040 "precr.qb.ph %[tmp_t8], %[tmp_t7], %[tmp_t8] \n"
1041 "sw %[tmp_t8], 0(%[dst_argb4444]) \n"
1042 ".set pop \n"
1043 :[tmp_t1] "=&r" (tmp_t1), [tmp_t2] "=&r" (tmp_t2),
1044 [tmp_t3] "=&r" (tmp_t3), [tmp_t4] "=&r" (tmp_t4),
1045 [tmp_t5] "=&r" (tmp_t5), [tmp_t6] "=&r" (tmp_t6),
1046 [tmp_t7] "=&r" (tmp_t7), [tmp_t8] "=&r" (tmp_t8),
1047 [tmp_t9] "=&r" (tmp_t9)
1048 :[dst_argb4444] "r" (dst_argb4444), [yg] "r" (yg),
1049 [src_u] "r" (src_u), [src_v] "r" (src_v),
1050 [src_y] "r" (src_y),
1051 [tmp_ub] "r" (tmp_ub), [tmp_ug] "r" (tmp_ug),
1052 [tmp_vg] "r" (tmp_vg), [tmp_vr] "r" (tmp_vr),
1053 [tmp_bb] "r" (tmp_bb), [tmp_bg] "r" (tmp_bg),
1054 [tmp_br] "r" (tmp_br), [tmp_yg] "r" (tmp_yg),
1055 [tmp_mask] "r" (tmp_mask)
1056 );
1057 src_y += 2;
1058 src_u += 1;
1059 src_v += 1;
1060 dst_argb4444 += 4; // Advance 2 pixels.
1061 }
1062 }
1063
1064 void I422ToARGB1555Row_DSPR2(const uint8* src_y,
1065 const uint8* src_u,
1066 const uint8* src_v,
1067 uint8* dst_argb1555,
1068 const struct YuvConstants* yuvconstants,
1069 int width) {
1070 int x;
1071 uint32 tmp_ub = yuvconstants->kUVToB[0];
1072 uint32 tmp_ug = yuvconstants->kUVToG[0];
1073 uint32 tmp_vg = yuvconstants->kUVToG[1];
1074 uint32 tmp_vr = yuvconstants->kUVToR[1];
1075 uint32 tmp_bb = yuvconstants->kUVBiasB[0];
1076 uint32 tmp_bg = yuvconstants->kUVBiasG[0];
1077 uint32 tmp_br = yuvconstants->kUVBiasR[0];
1078 uint32 yg = yuvconstants->kYToRgb[0];
1079 uint32 tmp_yg;
1080 uint32 tmp_mask = 0x80008000;
1081 tmp_bb = ((uint)(tmp_bb & 0xffff) << 16) | (tmp_bb & 0xffff);
1082 tmp_bg = ((uint)(tmp_bg & 0xffff) << 16) | (tmp_bg & 0xffff);
1083 tmp_br = ((uint)(tmp_br & 0xffff) << 16) | (tmp_br & 0xffff);
1084 tmp_yg = ((uint)(yg & 0xffff) << 16) | (yg & 0xffff);
1085 tmp_ub = ~(((uint)(tmp_ub & 0xffff) << 16) | (tmp_ub & 0xffff)) + 0x00010001;
1086 tmp_ug = ((uint)(tmp_ug & 0xffff) << 16) | (tmp_ug & 0xffff);
1087 tmp_vg = ((uint)(tmp_vg & 0xffff) << 16) | (tmp_vg & 0xffff);
1088 tmp_vr = ~(((uint)(tmp_vr & 0xffff) << 16) | (tmp_vr & 0xffff)) + 0x00010001;
1089 yg = yg * 0x0101;
1090
1091 for (x = 0; x < width - 1; x += 2) {
1092 uint32 tmp_t1, tmp_t2, tmp_t3, tmp_t4, tmp_t5;
1093 uint32 tmp_t6, tmp_t7, tmp_t8, tmp_t9;
1094 __asm__ __volatile__ (
1095 ".set push \n"
1096 ".set noreorder \n"
1097 "lbu %[tmp_t7], 0(%[src_y]) \n"
1098 "lbu %[tmp_t1], 1(%[src_y]) \n"
1099 "mul %[tmp_t7], %[tmp_t7], %[yg] \n"
1100 "mul %[tmp_t1], %[tmp_t1], %[yg] \n"
1101 "lbu %[tmp_t2], 0(%[src_u]) \n"
1102 "lbu %[tmp_t3], 0(%[src_v]) \n"
1103 "replv.ph %[tmp_t2], %[tmp_t2] \n"
1104 "replv.ph %[tmp_t3], %[tmp_t3] \n"
1105 "mul.ph %[tmp_t4], %[tmp_t2], %[tmp_ub] \n"
1106 "mul.ph %[tmp_t5], %[tmp_t2], %[tmp_ug] \n"
1107 "mul.ph %[tmp_t6], %[tmp_t3], %[tmp_vr] \n"
1108 "mul.ph %[tmp_t3], %[tmp_t3], %[tmp_vg] \n"
1109 "srl %[tmp_t7], %[tmp_t7], 16 \n"
1110 "ins %[tmp_t1], %[tmp_t7], 0, 16 \n"
1111 "addq_s.ph %[tmp_t7], %[tmp_t1], %[tmp_bb] \n"
1112 "addq_s.ph %[tmp_t8], %[tmp_t1], %[tmp_bg] \n"
1113 "addq_s.ph %[tmp_t9], %[tmp_t1], %[tmp_br] \n"
1114 "addq_s.ph %[tmp_t5], %[tmp_t5], %[tmp_t3] \n"
1115 "addq_s.ph %[tmp_t7], %[tmp_t7], %[tmp_t4] \n"
1116 "subq_s.ph %[tmp_t8], %[tmp_t8], %[tmp_t5] \n"
1117 "addq_s.ph %[tmp_t9], %[tmp_t9], %[tmp_t6] \n"
1118 "shra.ph %[tmp_t7], %[tmp_t7], 6 \n"
1119 "shra.ph %[tmp_t8], %[tmp_t8], 6 \n"
1120 "shra.ph %[tmp_t9], %[tmp_t9], 6 \n"
1121 "shll_s.ph %[tmp_t7], %[tmp_t7], 7 \n"
1122 "shll_s.ph %[tmp_t8], %[tmp_t8], 7 \n"
1123 "shll_s.ph %[tmp_t9], %[tmp_t9], 7 \n"
1124 "precrqu_s.qb.ph %[tmp_t8], %[tmp_mask], %[tmp_t8] \n"
1125 "precrqu_s.qb.ph %[tmp_t7], %[tmp_t9], %[tmp_t7] \n"
1126 "precrq.ph.w %[tmp_t2], %[tmp_t8], %[tmp_t7] \n"
1127 "ins %[tmp_t7], %[tmp_t8], 16, 16 \n"
1128 "precr.qb.ph %[tmp_t8], %[tmp_t2], %[tmp_t7] \n"
1129 "precrq.qb.ph %[tmp_t7], %[tmp_t2], %[tmp_t7] \n"
1130 "ins %[tmp_t3], %[tmp_t8], 7, 24 \n"
1131 "ins %[tmp_t3], %[tmp_t8], 10, 16 \n"
1132 "ins %[tmp_t3], %[tmp_t8], 13, 8 \n"
1133 "ins %[tmp_t4], %[tmp_t7], 7, 24 \n"
1134 "ins %[tmp_t4], %[tmp_t7], 10, 16 \n"
1135 "ins %[tmp_t4], %[tmp_t7], 13, 8 \n"
1136 "precrq.ph.w %[tmp_t8], %[tmp_t4], %[tmp_t3] \n"
1137 "or %[tmp_t8], %[tmp_t8], %[tmp_mask]\n"
1138 "sw %[tmp_t8], 0(%[dst_argb1555]) \n"
1139 ".set pop \n"
1140 :[tmp_t1] "=&r" (tmp_t1), [tmp_t2] "=&r" (tmp_t2),
1141 [tmp_t3] "=&r" (tmp_t3), [tmp_t4] "=&r" (tmp_t4),
1142 [tmp_t5] "=&r" (tmp_t5), [tmp_t6] "=&r" (tmp_t6),
1143 [tmp_t7] "=&r" (tmp_t7), [tmp_t8] "=&r" (tmp_t8),
1144 [tmp_t9] "=&r" (tmp_t9)
1145 :[dst_argb1555] "r" (dst_argb1555), [yg] "r" (yg),
1146 [src_u] "r" (src_u), [src_v] "r" (src_v),
1147 [src_y] "r" (src_y),
1148 [tmp_ub] "r" (tmp_ub), [tmp_ug] "r" (tmp_ug),
1149 [tmp_vg] "r" (tmp_vg), [tmp_vr] "r" (tmp_vr),
1150 [tmp_bb] "r" (tmp_bb), [tmp_bg] "r" (tmp_bg),
1151 [tmp_br] "r" (tmp_br), [tmp_yg] "r" (tmp_yg),
1152 [tmp_mask] "r" (tmp_mask)
1153 );
1154 src_y += 2;
1155 src_u += 1;
1156 src_v += 1;
1157 dst_argb1555 += 4; // Advance 2 pixels.
1158 }
1159 }
1160
1161 void NV12ToARGBRow_DSPR2(const uint8* src_y,
1162 const uint8* src_uv,
1163 uint8* rgb_buf,
1164 const struct YuvConstants* yuvconstants,
1165 int width) {
1166 int x;
1167 uint32 tmp_ub = yuvconstants->kUVToB[0];
1168 uint32 tmp_ug = yuvconstants->kUVToG[0];
1169 uint32 tmp_vg = yuvconstants->kUVToG[1];
1170 uint32 tmp_vr = yuvconstants->kUVToR[1];
1171 uint32 tmp_bb = yuvconstants->kUVBiasB[0];
1172 uint32 tmp_bg = yuvconstants->kUVBiasG[0];
1173 uint32 tmp_br = yuvconstants->kUVBiasR[0];
1174 uint32 yg = yuvconstants->kYToRgb[0];
1175 uint32 tmp_mask = 0x7fff7fff;
1176 uint32 tmp_yg;
1177 tmp_bb = ((uint)(tmp_bb & 0xffff) << 16) | (tmp_bb & 0xffff);
1178 tmp_bg = ((uint)(tmp_bg & 0xffff) << 16) | (tmp_bg & 0xffff);
1179 tmp_br = ((uint)(tmp_br & 0xffff) << 16) | (tmp_br & 0xffff);
1180 tmp_yg = ((uint)(yg & 0xffff) << 16) | (yg & 0xffff);
1181 tmp_ub = ~(((uint)(tmp_ub & 0xffff) << 16) | (tmp_ub & 0xffff)) + 0x00010001;
1182 tmp_ug = ((uint)(tmp_ug & 0xffff) << 16) | (tmp_ug & 0xffff);
1183 tmp_vg = ((uint)(tmp_vg & 0xffff) << 16) | (tmp_vg & 0xffff);
1184 tmp_vr = ~(((uint)(tmp_vr & 0xffff) << 16) | (tmp_vr & 0xffff)) + 0x00010001;
1185 yg = yg * 0x0101;
1186
1187 for (x = 0; x < width - 1; x += 2) {
1188 uint32 tmp_t1, tmp_t2, tmp_t3, tmp_t4, tmp_t5;
1189 uint32 tmp_t6, tmp_t7, tmp_t8, tmp_t9;
1190 __asm__ __volatile__ (
1191 ".set push \n"
1192 ".set noreorder \n"
1193 "lbu %[tmp_t7], 0(%[src_y]) \n"
1194 "lbu %[tmp_t1], 1(%[src_y]) \n"
1195 "mul %[tmp_t7], %[tmp_t7], %[yg] \n"
1196 "mul %[tmp_t1], %[tmp_t1], %[yg] \n"
1197 "lbu %[tmp_t2], 0(%[src_uv]) \n"
1198 "lbu %[tmp_t3], 1(%[src_uv]) \n"
1199 "replv.ph %[tmp_t2], %[tmp_t2] \n"
1200 "replv.ph %[tmp_t3], %[tmp_t3] \n"
1201 "mul.ph %[tmp_t4], %[tmp_t2], %[tmp_ub] \n"
1202 "mul.ph %[tmp_t5], %[tmp_t2], %[tmp_ug] \n"
1203 "mul.ph %[tmp_t6], %[tmp_t3], %[tmp_vr] \n"
1204 "mul.ph %[tmp_t3], %[tmp_t3], %[tmp_vg] \n"
1205 "srl %[tmp_t7], %[tmp_t7], 16 \n"
1206 "ins %[tmp_t1], %[tmp_t7], 0, 16 \n"
1207 "addq_s.ph %[tmp_t7], %[tmp_t1], %[tmp_bb] \n"
1208 "addq_s.ph %[tmp_t8], %[tmp_t1], %[tmp_bg] \n"
1209 "addq_s.ph %[tmp_t9], %[tmp_t1], %[tmp_br] \n"
1210 "addq_s.ph %[tmp_t5], %[tmp_t5], %[tmp_t3] \n"
1211 "addq_s.ph %[tmp_t7], %[tmp_t7], %[tmp_t4] \n"
1212 "subq_s.ph %[tmp_t8], %[tmp_t8], %[tmp_t5] \n"
1213 "addq_s.ph %[tmp_t9], %[tmp_t9], %[tmp_t6] \n"
1214 "shra.ph %[tmp_t7], %[tmp_t7], 6 \n"
1215 "shra.ph %[tmp_t8], %[tmp_t8], 6 \n"
1216 "shra.ph %[tmp_t9], %[tmp_t9], 6 \n"
1217 "shll_s.ph %[tmp_t7], %[tmp_t7], 7 \n"
1218 "shll_s.ph %[tmp_t8], %[tmp_t8], 7 \n"
1219 "shll_s.ph %[tmp_t9], %[tmp_t9], 7 \n"
1220 "precrqu_s.qb.ph %[tmp_t8], %[tmp_mask], %[tmp_t8] \n"
1221 "precrqu_s.qb.ph %[tmp_t7], %[tmp_t9], %[tmp_t7] \n"
1222 "precrq.ph.w %[tmp_t2], %[tmp_t8], %[tmp_t7] \n"
1223 "ins %[tmp_t7], %[tmp_t8], 16, 16 \n"
1224 "precr.qb.ph %[tmp_t8], %[tmp_t2], %[tmp_t7] \n"
1225 "precrq.qb.ph %[tmp_t7], %[tmp_t2], %[tmp_t7] \n"
1226 "sw %[tmp_t8], 0(%[rgb_buf]) \n"
1227 "sw %[tmp_t7], 4(%[rgb_buf]) \n"
1228 ".set pop \n"
1229 :[tmp_t1] "=&r" (tmp_t1), [tmp_t2] "=&r" (tmp_t2),
1230 [tmp_t3] "=&r" (tmp_t3), [tmp_t4] "=&r" (tmp_t4),
1231 [tmp_t5] "=&r" (tmp_t5), [tmp_t6] "=&r" (tmp_t6),
1232 [tmp_t7] "=&r" (tmp_t7), [tmp_t8] "=&r" (tmp_t8),
1233 [tmp_t9] "=&r" (tmp_t9)
1234 :[src_y] "r" (src_y), [src_uv] "r" (src_uv), [yg] "r" (yg),
1235 [tmp_ub] "r" (tmp_ub), [tmp_ug] "r" (tmp_ug),
1236 [tmp_vg] "r" (tmp_vg), [tmp_vr] "r" (tmp_vr),
1237 [tmp_bb] "r" (tmp_bb), [tmp_bg] "r" (tmp_bg),
1238 [tmp_br] "r" (tmp_br), [tmp_yg] "r" (tmp_yg),
1239 [rgb_buf] "r" (rgb_buf), [tmp_mask] "r" (tmp_mask)
1240 );
1241
1242 src_y += 2;
1243 src_uv += 2;
1244 rgb_buf += 8; // Advance 2 pixels.
1245 }
1246 }
1247
1248 void BGRAToUVRow_DSPR2(const uint8* src_rgb0, int src_stride_rgb,
1249 uint8* dst_u, uint8* dst_v, int width) {
1250 const uint8* src_rgb1 = src_rgb0 + src_stride_rgb;
1251 int x;
1252 int const1 = 0xffda0000;
1253 int const2 = 0x0070ffb6;
1254 int const3 = 0x00700000;
1255 int const4 = 0xffeeffa2;
1256 int const5 = 0x100;
1257 for (x = 0; x < width - 1; x += 2) {
1258 int tmp_t1, tmp_t2, tmp_t3, tmp_t4, tmp_t5;
1259 int tmp_t6, tmp_t7, tmp_t8;
1260 __asm__ __volatile__ (
1261 ".set push \n"
1262 ".set noreorder \n"
1263 "lw %[tmp_t1], 0(%[src_rgb0]) \n"
1264 "lw %[tmp_t2], 4(%[src_rgb0]) \n"
1265 "lw %[tmp_t3], 0(%[src_rgb1]) \n"
1266 "lw %[tmp_t4], 4(%[src_rgb1]) \n"
1267 "preceu.ph.qbr %[tmp_t5], %[tmp_t1] \n"
1268 "preceu.ph.qbl %[tmp_t1], %[tmp_t1] \n"
1269 "preceu.ph.qbr %[tmp_t6], %[tmp_t2] \n"
1270 "preceu.ph.qbl %[tmp_t2], %[tmp_t2] \n"
1271 "preceu.ph.qbr %[tmp_t7], %[tmp_t3] \n"
1272 "preceu.ph.qbl %[tmp_t3], %[tmp_t3] \n"
1273 "preceu.ph.qbr %[tmp_t8], %[tmp_t4] \n"
1274 "preceu.ph.qbl %[tmp_t4], %[tmp_t4] \n"
1275 "addu.ph %[tmp_t5], %[tmp_t5], %[tmp_t6] \n"
1276 "addu.ph %[tmp_t7], %[tmp_t7], %[tmp_t8] \n"
1277 "addu.ph %[tmp_t1], %[tmp_t1], %[tmp_t2] \n"
1278 "addu.ph %[tmp_t3], %[tmp_t3], %[tmp_t4] \n"
1279 "addu.ph %[tmp_t5], %[tmp_t5], %[tmp_t7] \n"
1280 "addu.ph %[tmp_t1], %[tmp_t1], %[tmp_t3] \n"
1281 "shrl.ph %[tmp_t5], %[tmp_t5], 2 \n"
1282 "shrl.ph %[tmp_t1], %[tmp_t1], 2 \n"
1283 "mult $ac0, %[const5], %[const5] \n"
1284 "mult $ac1, %[const5], %[const5] \n"
1285 "dpaq_s.w.ph $ac0, %[tmp_t5], %[const1] \n"
1286 "dpaq_s.w.ph $ac1, %[tmp_t5], %[const3] \n"
1287 "dpaq_s.w.ph $ac0, %[tmp_t1], %[const2] \n"
1288 "dpaq_s.w.ph $ac1, %[tmp_t1], %[const4] \n"
1289 "extr_r.w %[tmp_t7], $ac0, 9 \n"
1290 "extr_r.w %[tmp_t8], $ac1, 9 \n"
1291 "addiu %[dst_u], %[dst_u], 1 \n"
1292 "addiu %[dst_v], %[dst_v], 1 \n"
1293 "addiu %[src_rgb0], %[src_rgb0], 8 \n"
1294 "addiu %[src_rgb1], %[src_rgb1], 8 \n"
1295 "sb %[tmp_t7], -1(%[dst_u]) \n"
1296 "sb %[tmp_t8], -1(%[dst_v]) \n"
1297 ".set pop \n"
1298 :[tmp_t1] "=&r" (tmp_t1), [tmp_t2] "=&r" (tmp_t2),
1299 [tmp_t3] "=&r" (tmp_t3), [tmp_t4] "=&r" (tmp_t4),
1300 [tmp_t5] "=&r" (tmp_t5), [tmp_t6] "=&r" (tmp_t6),
1301 [tmp_t7] "=&r" (tmp_t7), [tmp_t8] "=&r" (tmp_t8),
1302 [src_rgb0] "+r" (src_rgb0), [src_rgb1] "+r" (src_rgb1),
1303 [dst_u] "+r" (dst_u), [dst_v] "+r" (dst_v)
1304 :[const1] "r" (const1), [const2] "r" (const2),
1305 [const3] "r" (const3), [const4] "r" (const4),
1306 [const5] "r" (const5)
1307 :"hi", "lo" ,"$ac1lo", "$ac1hi"
1308 );
1309 }
1310 }
1311
1312 void BGRAToYRow_DSPR2(const uint8* src_argb0, uint8* dst_y, int width) {
1313 int x;
1314 int const1 = 0x00420000;
1315 int const2 = 0x00190081;
1316 int const5 = 0x40;
1317 for (x = 0; x < width; x+=4) {
1318 int tmp_t1, tmp_t2, tmp_t3, tmp_t4, tmp_t5;
1319 int tmp_t6, tmp_t7, tmp_t8;
1320 __asm__ __volatile__ (
1321 ".set push \n"
1322 ".set noreorder \n"
1323 "lw %[tmp_t1], 0(%[src_argb0]) \n"
1324 "lw %[tmp_t2], 4(%[src_argb0]) \n"
1325 "lw %[tmp_t3], 8(%[src_argb0]) \n"
1326 "lw %[tmp_t4], 12(%[src_argb0]) \n"
1327 "preceu.ph.qbr %[tmp_t5], %[tmp_t1] \n"
1328 "preceu.ph.qbl %[tmp_t1], %[tmp_t1] \n"
1329 "preceu.ph.qbr %[tmp_t6], %[tmp_t2] \n"
1330 "preceu.ph.qbl %[tmp_t2], %[tmp_t2] \n"
1331 "preceu.ph.qbr %[tmp_t7], %[tmp_t3] \n"
1332 "preceu.ph.qbl %[tmp_t3], %[tmp_t3] \n"
1333 "preceu.ph.qbr %[tmp_t8], %[tmp_t4] \n"
1334 "preceu.ph.qbl %[tmp_t4], %[tmp_t4] \n"
1335 "mult $ac0, %[const5], %[const5] \n"
1336 "mult $ac1, %[const5], %[const5] \n"
1337 "mult $ac2, %[const5], %[const5] \n"
1338 "mult $ac3, %[const5], %[const5] \n"
1339 "dpa.w.ph $ac0, %[tmp_t5], %[const1] \n"
1340 "dpa.w.ph $ac1, %[tmp_t6], %[const1] \n"
1341 "dpa.w.ph $ac2, %[tmp_t7], %[const1] \n"
1342 "dpa.w.ph $ac3, %[tmp_t8], %[const1] \n"
1343 "dpa.w.ph $ac0, %[tmp_t1], %[const2] \n"
1344 "dpa.w.ph $ac1, %[tmp_t2], %[const2] \n"
1345 "dpa.w.ph $ac2, %[tmp_t3], %[const2] \n"
1346 "dpa.w.ph $ac3, %[tmp_t4], %[const2] \n"
1347 "extr_r.w %[tmp_t1], $ac0, 8 \n"
1348 "extr_r.w %[tmp_t2], $ac1, 8 \n"
1349 "extr_r.w %[tmp_t3], $ac2, 8 \n"
1350 "extr_r.w %[tmp_t4], $ac3, 8 \n"
1351 "addiu %[src_argb0],%[src_argb0], 16 \n"
1352 "addiu %[dst_y], %[dst_y], 4 \n"
1353 "sb %[tmp_t1], -4(%[dst_y]) \n"
1354 "sb %[tmp_t2], -3(%[dst_y]) \n"
1355 "sb %[tmp_t3], -2(%[dst_y]) \n"
1356 "sb %[tmp_t4], -1(%[dst_y]) \n"
1357 ".set pop \n"
1358 :[tmp_t1] "=&r" (tmp_t1), [tmp_t2] "=&r" (tmp_t2),
1359 [tmp_t3] "=&r" (tmp_t3), [tmp_t4] "=&r" (tmp_t4),
1360 [tmp_t5] "=&r" (tmp_t5), [tmp_t6] "=&r" (tmp_t6),
1361 [tmp_t7] "=&r" (tmp_t7), [tmp_t8] "=&r" (tmp_t8),
1362 [src_argb0] "+r" (src_argb0), [dst_y] "+r" (dst_y)
1363 :[const1] "r" (const1), [const2] "r" (const2),
1364 [const5] "r" (const5)
1365 :"hi", "lo" , "$ac1lo", "$ac1hi", "$ac2lo", "$ac2hi",
1366 "$ac3lo", "$ac3hi"
1367 );
1368 }
1369 }
1370
1371 void ABGRToUVRow_DSPR2(const uint8* src_rgb0, int src_stride_rgb,
1372 uint8* dst_u, uint8* dst_v, int width) {
1373 const uint8* src_rgb1 = src_rgb0 + src_stride_rgb;
1374 int x;
1375 int const1 = 0xffb6ffda;
1376 int const2 = 0x00000070;
1377 int const3 = 0xffa20070;
1378 int const4 = 0x0000ffee;
1379 int const5 = 0x100;
1380
1381 for (x = 0; x < width - 1; x += 2) {
1382 int tmp_t1, tmp_t2, tmp_t3, tmp_t4, tmp_t5;
1383 int tmp_t6, tmp_t7, tmp_t8;
1384 __asm__ __volatile__ (
1385 ".set push \n"
1386 ".set noreorder \n"
1387 "lw %[tmp_t1], 0(%[src_rgb0]) \n"
1388 "lw %[tmp_t2], 4(%[src_rgb0]) \n"
1389 "lw %[tmp_t3], 0(%[src_rgb1]) \n"
1390 "lw %[tmp_t4], 4(%[src_rgb1]) \n"
1391 "preceu.ph.qbr %[tmp_t5], %[tmp_t1] \n"
1392 "preceu.ph.qbl %[tmp_t1], %[tmp_t1] \n"
1393 "preceu.ph.qbr %[tmp_t6], %[tmp_t2] \n"
1394 "preceu.ph.qbl %[tmp_t2], %[tmp_t2] \n"
1395 "preceu.ph.qbr %[tmp_t7], %[tmp_t3] \n"
1396 "preceu.ph.qbl %[tmp_t3], %[tmp_t3] \n"
1397 "preceu.ph.qbr %[tmp_t8], %[tmp_t4] \n"
1398 "preceu.ph.qbl %[tmp_t4], %[tmp_t4] \n"
1399 "addu.ph %[tmp_t5], %[tmp_t5], %[tmp_t6] \n"
1400 "addu.ph %[tmp_t7], %[tmp_t7], %[tmp_t8] \n"
1401 "addu.ph %[tmp_t1], %[tmp_t1], %[tmp_t2] \n"
1402 "addu.ph %[tmp_t3], %[tmp_t3], %[tmp_t4] \n"
1403 "addu.ph %[tmp_t5], %[tmp_t5], %[tmp_t7] \n"
1404 "addu.ph %[tmp_t1], %[tmp_t1], %[tmp_t3] \n"
1405 "shrl.ph %[tmp_t5], %[tmp_t5], 2 \n"
1406 "shrl.ph %[tmp_t1], %[tmp_t1], 2 \n"
1407 "mult $ac0, %[const5], %[const5] \n"
1408 "mult $ac1, %[const5], %[const5] \n"
1409 "dpaq_s.w.ph $ac0, %[tmp_t5], %[const1] \n"
1410 "dpaq_s.w.ph $ac1, %[tmp_t5], %[const3] \n"
1411 "dpaq_s.w.ph $ac0, %[tmp_t1], %[const2] \n"
1412 "dpaq_s.w.ph $ac1, %[tmp_t1], %[const4] \n"
1413 "extr_r.w %[tmp_t7], $ac0, 9 \n"
1414 "extr_r.w %[tmp_t8], $ac1, 9 \n"
1415 "addiu %[dst_u], %[dst_u], 1 \n"
1416 "addiu %[dst_v], %[dst_v], 1 \n"
1417 "addiu %[src_rgb0], %[src_rgb0], 8 \n"
1418 "addiu %[src_rgb1], %[src_rgb1], 8 \n"
1419 "sb %[tmp_t7], -1(%[dst_u]) \n"
1420 "sb %[tmp_t8], -1(%[dst_v]) \n"
1421 ".set pop \n"
1422 :[tmp_t1] "=&r" (tmp_t1), [tmp_t2] "=&r" (tmp_t2),
1423 [tmp_t3] "=&r" (tmp_t3), [tmp_t4] "=&r" (tmp_t4),
1424 [tmp_t5] "=&r" (tmp_t5), [tmp_t6] "=&r" (tmp_t6),
1425 [tmp_t7] "=&r" (tmp_t7), [tmp_t8] "=&r" (tmp_t8),
1426 [src_rgb0] "+r" (src_rgb0), [src_rgb1] "+r" (src_rgb1),
1427 [dst_u] "+r" (dst_u), [dst_v] "+r" (dst_v)
1428 :[const1] "r" (const1), [const2] "r" (const2),
1429 [const3] "r" (const3), [const4] "r" (const4),
1430 [const5] "r" (const5)
1431 :"hi", "lo" ,"$ac1lo", "$ac1hi"
1432 );
1433 }
1434 }
1435
1436 void ARGBToYRow_DSPR2(const uint8* src_argb0, uint8* dst_y, int width) {
1437 int x;
1438 int const1 = 0x00810019;
1439 int const2 = 0x00000042;
1440 int const5 = 0x40;
1441 for (x = 0; x < width; x+=4) {
1442 int tmp_t1, tmp_t2, tmp_t3, tmp_t4, tmp_t5;
1443 int tmp_t6, tmp_t7, tmp_t8;
1444 __asm__ __volatile__ (
1445 ".set push \n"
1446 ".set noreorder \n"
1447 "lw %[tmp_t1], 0(%[src_argb0]) \n"
1448 "lw %[tmp_t2], 4(%[src_argb0]) \n"
1449 "lw %[tmp_t3], 8(%[src_argb0]) \n"
1450 "lw %[tmp_t4], 12(%[src_argb0]) \n"
1451 "preceu.ph.qbr %[tmp_t5], %[tmp_t1] \n"
1452 "preceu.ph.qbl %[tmp_t1], %[tmp_t1] \n"
1453 "preceu.ph.qbr %[tmp_t6], %[tmp_t2] \n"
1454 "preceu.ph.qbl %[tmp_t2], %[tmp_t2] \n"
1455 "preceu.ph.qbr %[tmp_t7], %[tmp_t3] \n"
1456 "preceu.ph.qbl %[tmp_t3], %[tmp_t3] \n"
1457 "preceu.ph.qbr %[tmp_t8], %[tmp_t4] \n"
1458 "preceu.ph.qbl %[tmp_t4], %[tmp_t4] \n"
1459 "mult $ac0, %[const5], %[const5] \n"
1460 "mult $ac1, %[const5], %[const5] \n"
1461 "mult $ac2, %[const5], %[const5] \n"
1462 "mult $ac3, %[const5], %[const5] \n"
1463 "dpa.w.ph $ac0, %[tmp_t5], %[const1] \n"
1464 "dpa.w.ph $ac1, %[tmp_t6], %[const1] \n"
1465 "dpa.w.ph $ac2, %[tmp_t7], %[const1] \n"
1466 "dpa.w.ph $ac3, %[tmp_t8], %[const1] \n"
1467 "dpa.w.ph $ac0, %[tmp_t1], %[const2] \n"
1468 "dpa.w.ph $ac1, %[tmp_t2], %[const2] \n"
1469 "dpa.w.ph $ac2, %[tmp_t3], %[const2] \n"
1470 "dpa.w.ph $ac3, %[tmp_t4], %[const2] \n"
1471 "extr_r.w %[tmp_t1], $ac0, 8 \n"
1472 "extr_r.w %[tmp_t2], $ac1, 8 \n"
1473 "extr_r.w %[tmp_t3], $ac2, 8 \n"
1474 "extr_r.w %[tmp_t4], $ac3, 8 \n"
1475 "addiu %[dst_y], %[dst_y], 4 \n"
1476 "addiu %[src_argb0],%[src_argb0], 16 \n"
1477 "sb %[tmp_t1], -4(%[dst_y]) \n"
1478 "sb %[tmp_t2], -3(%[dst_y]) \n"
1479 "sb %[tmp_t3], -2(%[dst_y]) \n"
1480 "sb %[tmp_t4], -1(%[dst_y]) \n"
1481 ".set pop \n"
1482 :[tmp_t1] "=&r" (tmp_t1), [tmp_t2] "=&r" (tmp_t2),
1483 [tmp_t3] "=&r" (tmp_t3), [tmp_t4] "=&r" (tmp_t4),
1484 [tmp_t5] "=&r" (tmp_t5), [tmp_t6] "=&r" (tmp_t6),
1485 [tmp_t7] "=&r" (tmp_t7), [tmp_t8] "=&r" (tmp_t8),
1486 [src_argb0] "+r" (src_argb0), [dst_y] "+r" (dst_y)
1487 :[const1] "r" (const1), [const2] "r" (const2),
1488 [const5] "r" (const5)
1489 :"hi", "lo" , "$ac1lo", "$ac1hi", "$ac2lo", "$ac2hi",
1490 "$ac3lo", "$ac3hi"
1491 );
1492 }
1493 }
1494
1495 void ABGRToYRow_DSPR2(const uint8* src_argb0, uint8* dst_y, int width) {
1496 int x;
1497 int const1 = 0x00810042;
1498 int const2 = 0x00000019;
1499 int const5 = 0x40;
1500 for (x = 0; x < width; x+=4) {
1501 int tmp_t1, tmp_t2, tmp_t3, tmp_t4, tmp_t5;
1502 int tmp_t6, tmp_t7, tmp_t8;
1503 __asm__ __volatile__ (
1504 ".set push \n"
1505 ".set noreorder \n"
1506 "lw %[tmp_t1], 0(%[src_argb0]) \n"
1507 "lw %[tmp_t2], 4(%[src_argb0]) \n"
1508 "lw %[tmp_t3], 8(%[src_argb0]) \n"
1509 "lw %[tmp_t4], 12(%[src_argb0]) \n"
1510 "preceu.ph.qbr %[tmp_t5], %[tmp_t1] \n"
1511 "preceu.ph.qbl %[tmp_t1], %[tmp_t1] \n"
1512 "preceu.ph.qbr %[tmp_t6], %[tmp_t2] \n"
1513 "preceu.ph.qbl %[tmp_t2], %[tmp_t2] \n"
1514 "preceu.ph.qbr %[tmp_t7], %[tmp_t3] \n"
1515 "preceu.ph.qbl %[tmp_t3], %[tmp_t3] \n"
1516 "preceu.ph.qbr %[tmp_t8], %[tmp_t4] \n"
1517 "preceu.ph.qbl %[tmp_t4], %[tmp_t4] \n"
1518 "mult $ac0, %[const5], %[const5] \n"
1519 "mult $ac1, %[const5], %[const5] \n"
1520 "mult $ac2, %[const5], %[const5] \n"
1521 "mult $ac3, %[const5], %[const5] \n"
1522 "dpa.w.ph $ac0, %[tmp_t5], %[const1] \n"
1523 "dpa.w.ph $ac1, %[tmp_t6], %[const1] \n"
1524 "dpa.w.ph $ac2, %[tmp_t7], %[const1] \n"
1525 "dpa.w.ph $ac3, %[tmp_t8], %[const1] \n"
1526 "dpa.w.ph $ac0, %[tmp_t1], %[const2] \n"
1527 "dpa.w.ph $ac1, %[tmp_t2], %[const2] \n"
1528 "dpa.w.ph $ac2, %[tmp_t3], %[const2] \n"
1529 "dpa.w.ph $ac3, %[tmp_t4], %[const2] \n"
1530 "extr_r.w %[tmp_t1], $ac0, 8 \n"
1531 "extr_r.w %[tmp_t2], $ac1, 8 \n"
1532 "extr_r.w %[tmp_t3], $ac2, 8 \n"
1533 "extr_r.w %[tmp_t4], $ac3, 8 \n"
1534 "addiu %[src_argb0],%[src_argb0], 16 \n"
1535 "addiu %[dst_y], %[dst_y], 4 \n"
1536 "sb %[tmp_t1], -4(%[dst_y]) \n"
1537 "sb %[tmp_t2], -3(%[dst_y]) \n"
1538 "sb %[tmp_t3], -2(%[dst_y]) \n"
1539 "sb %[tmp_t4], -1(%[dst_y]) \n"
1540 ".set pop \n"
1541 :[tmp_t1] "=&r" (tmp_t1), [tmp_t2] "=&r" (tmp_t2),
1542 [tmp_t3] "=&r" (tmp_t3), [tmp_t4] "=&r" (tmp_t4),
1543 [tmp_t5] "=&r" (tmp_t5), [tmp_t6] "=&r" (tmp_t6),
1544 [tmp_t7] "=&r" (tmp_t7), [tmp_t8] "=&r" (tmp_t8),
1545 [src_argb0] "+r" (src_argb0), [dst_y] "+r" (dst_y)
1546 :[const1] "r" (const1), [const2] "r" (const2),
1547 [const5] "r" (const5)
1548 :"hi", "lo" , "$ac1lo", "$ac1hi", "$ac2lo", "$ac2hi",
1549 "$ac3lo", "$ac3hi"
1550 );
1551 }
1552 }
1553
1554 void RGBAToUVRow_DSPR2(const uint8* src_rgb0, int src_stride_rgb,
1555 uint8* dst_u, uint8* dst_v, int width) {
1556 const uint8* src_rgb1 = src_rgb0 + src_stride_rgb;
1557 int x;
1558 int const1 = 0xffb60070;
1559 int const2 = 0x0000ffda;
1560 int const3 = 0xffa2ffee;
1561 int const4 = 0x00000070;
1562 int const5 = 0x100;
1563
1564 for (x = 0; x < width - 1; x += 2) {
1565 int tmp_t1, tmp_t2, tmp_t3, tmp_t4, tmp_t5;
1566 int tmp_t6, tmp_t7, tmp_t8;
1567 __asm__ __volatile__ (
1568 ".set push \n"
1569 ".set noreorder \n"
1570 "ulw %[tmp_t1], 0+1(%[src_rgb0]) \n"
1571 "ulw %[tmp_t2], 4+1(%[src_rgb0]) \n"
1572 "ulw %[tmp_t3], 0+1(%[src_rgb1]) \n"
1573 "ulw %[tmp_t4], 4+1(%[src_rgb1]) \n"
1574 "preceu.ph.qbr %[tmp_t5], %[tmp_t1] \n"
1575 "preceu.ph.qbl %[tmp_t1], %[tmp_t1] \n"
1576 "preceu.ph.qbr %[tmp_t6], %[tmp_t2] \n"
1577 "preceu.ph.qbl %[tmp_t2], %[tmp_t2] \n"
1578 "preceu.ph.qbr %[tmp_t7], %[tmp_t3] \n"
1579 "preceu.ph.qbl %[tmp_t3], %[tmp_t3] \n"
1580 "preceu.ph.qbr %[tmp_t8], %[tmp_t4] \n"
1581 "preceu.ph.qbl %[tmp_t4], %[tmp_t4] \n"
1582 "addu.ph %[tmp_t5], %[tmp_t5], %[tmp_t6] \n"
1583 "addu.ph %[tmp_t7], %[tmp_t7], %[tmp_t8] \n"
1584 "addu.ph %[tmp_t1], %[tmp_t1], %[tmp_t2] \n"
1585 "addu.ph %[tmp_t3], %[tmp_t3], %[tmp_t4] \n"
1586 "addu.ph %[tmp_t5], %[tmp_t5], %[tmp_t7] \n"
1587 "addu.ph %[tmp_t1], %[tmp_t1], %[tmp_t3] \n"
1588 "shrl.ph %[tmp_t5], %[tmp_t5], 2 \n"
1589 "shrl.ph %[tmp_t1], %[tmp_t1], 2 \n"
1590 "mult $ac0, %[const5], %[const5] \n"
1591 "mult $ac1, %[const5], %[const5] \n"
1592 "dpaq_s.w.ph $ac0, %[tmp_t5], %[const1] \n"
1593 "dpaq_s.w.ph $ac1, %[tmp_t5], %[const3] \n"
1594 "dpaq_s.w.ph $ac0, %[tmp_t1], %[const2] \n"
1595 "dpaq_s.w.ph $ac1, %[tmp_t1], %[const4] \n"
1596 "extr_r.w %[tmp_t7], $ac0, 9 \n"
1597 "extr_r.w %[tmp_t8], $ac1, 9 \n"
1598 "addiu %[src_rgb0], %[src_rgb0], 8 \n"
1599 "addiu %[src_rgb1], %[src_rgb1], 8 \n"
1600 "addiu %[dst_u], %[dst_u], 1 \n"
1601 "addiu %[dst_v], %[dst_v], 1 \n"
1602 "sb %[tmp_t7], -1(%[dst_u]) \n"
1603 "sb %[tmp_t8], -1(%[dst_v]) \n"
1604 ".set pop \n"
1605 :[tmp_t1] "=&r" (tmp_t1), [tmp_t2] "=&r" (tmp_t2),
1606 [tmp_t3] "=&r" (tmp_t3), [tmp_t4] "=&r" (tmp_t4),
1607 [tmp_t5] "=&r" (tmp_t5), [tmp_t6] "=&r" (tmp_t6),
1608 [tmp_t7] "=&r" (tmp_t7), [tmp_t8] "=&r" (tmp_t8),
1609 [src_rgb0] "+r" (src_rgb0), [src_rgb1] "+r" (src_rgb1),
1610 [dst_u] "+r" (dst_u), [dst_v] "+r" (dst_v)
1611 :[const1] "r" (const1), [const2] "r" (const2),
1612 [const3] "r" (const3), [const4] "r" (const4),
1613 [const5] "r" (const5)
1614 :"hi", "lo" ,"$ac1lo", "$ac1hi"
1615 );
1616 }
1617 }
1618
1619 void RGBAToYRow_DSPR2(const uint8* src_argb0, uint8* dst_y, int width) {
1620 int x;
1621 int const1 = 0x00420081;
1622 int const2 = 0x00190000;
1623 int const5 = 0x40;
1624 for (x = 0; x < width; x+=4) {
1625 int tmp_t1, tmp_t2, tmp_t3, tmp_t4, tmp_t5;
1626 int tmp_t6, tmp_t7, tmp_t8;
1627 __asm__ __volatile__ (
1628 ".set push \n"
1629 ".set noreorder \n"
1630 "lw %[tmp_t1], 0(%[src_argb0]) \n"
1631 "lw %[tmp_t2], 4(%[src_argb0]) \n"
1632 "lw %[tmp_t3], 8(%[src_argb0]) \n"
1633 "lw %[tmp_t4], 12(%[src_argb0]) \n"
1634 "preceu.ph.qbl %[tmp_t5], %[tmp_t1] \n"
1635 "preceu.ph.qbr %[tmp_t1], %[tmp_t1] \n"
1636 "preceu.ph.qbl %[tmp_t6], %[tmp_t2] \n"
1637 "preceu.ph.qbr %[tmp_t2], %[tmp_t2] \n"
1638 "preceu.ph.qbl %[tmp_t7], %[tmp_t3] \n"
1639 "preceu.ph.qbr %[tmp_t3], %[tmp_t3] \n"
1640 "preceu.ph.qbl %[tmp_t8], %[tmp_t4] \n"
1641 "preceu.ph.qbr %[tmp_t4], %[tmp_t4] \n"
1642 "mult $ac0, %[const5], %[const5] \n"
1643 "mult $ac1, %[const5], %[const5] \n"
1644 "mult $ac2, %[const5], %[const5] \n"
1645 "mult $ac3, %[const5], %[const5] \n"
1646 "dpa.w.ph $ac0, %[tmp_t5], %[const1] \n"
1647 "dpa.w.ph $ac1, %[tmp_t6], %[const1] \n"
1648 "dpa.w.ph $ac2, %[tmp_t7], %[const1] \n"
1649 "dpa.w.ph $ac3, %[tmp_t8], %[const1] \n"
1650 "dpa.w.ph $ac0, %[tmp_t1], %[const2] \n"
1651 "dpa.w.ph $ac1, %[tmp_t2], %[const2] \n"
1652 "dpa.w.ph $ac2, %[tmp_t3], %[const2] \n"
1653 "dpa.w.ph $ac3, %[tmp_t4], %[const2] \n"
1654 "extr_r.w %[tmp_t1], $ac0, 8 \n"
1655 "extr_r.w %[tmp_t2], $ac1, 8 \n"
1656 "extr_r.w %[tmp_t3], $ac2, 8 \n"
1657 "extr_r.w %[tmp_t4], $ac3, 8 \n"
1658 "addiu %[dst_y], %[dst_y], 4 \n"
1659 "addiu %[src_argb0],%[src_argb0], 16 \n"
1660 "sb %[tmp_t1], -4(%[dst_y]) \n"
1661 "sb %[tmp_t2], -3(%[dst_y]) \n"
1662 "sb %[tmp_t3], -2(%[dst_y]) \n"
1663 "sb %[tmp_t4], -1(%[dst_y]) \n"
1664 ".set pop \n"
1665 :[tmp_t1] "=&r" (tmp_t1), [tmp_t2] "=&r" (tmp_t2),
1666 [tmp_t3] "=&r" (tmp_t3), [tmp_t4] "=&r" (tmp_t4),
1667 [tmp_t5] "=&r" (tmp_t5), [tmp_t6] "=&r" (tmp_t6),
1668 [tmp_t7] "=&r" (tmp_t7), [tmp_t8] "=&r" (tmp_t8),
1669 [src_argb0] "+r" (src_argb0), [dst_y] "+r" (dst_y)
1670 :[const1] "r" (const1), [const2] "r" (const2),
1671 [const5] "r" (const5)
1672 :"hi", "lo" , "$ac1lo", "$ac1hi", "$ac2lo", "$ac2hi",
1673 "$ac3lo", "$ac3hi"
1674 );
1675 }
1676 }
1677
1678 void ARGBToUVRow_DSPR2(const uint8* src_rgb0, int src_stride_rgb,
1679 uint8* dst_u, uint8* dst_v, int width) {
1680 const uint8* src_rgb1 = src_rgb0 + src_stride_rgb;
1681 int x;
1682 int const1 = 0xffb60070;
1683 int const2 = 0x0000ffda;
1684 int const3 = 0xffa2ffee;
1685 int const4 = 0x00000070;
1686 int const5 = 0x100;
1687
1688 for (x = 0; x < width - 1; x += 2) {
1689 int tmp_t1, tmp_t2, tmp_t3, tmp_t4, tmp_t5;
1690 int tmp_t6, tmp_t7, tmp_t8;
1691 __asm__ __volatile__ (
1692 ".set push \n"
1693 ".set noreorder \n"
1694 "lw %[tmp_t1], 0(%[src_rgb0]) \n"
1695 "lw %[tmp_t2], 4(%[src_rgb0]) \n"
1696 "lw %[tmp_t3], 0(%[src_rgb1]) \n"
1697 "lw %[tmp_t4], 4(%[src_rgb1]) \n"
1698 "preceu.ph.qbr %[tmp_t5], %[tmp_t1] \n"
1699 "preceu.ph.qbl %[tmp_t1], %[tmp_t1] \n"
1700 "preceu.ph.qbr %[tmp_t6], %[tmp_t2] \n"
1701 "preceu.ph.qbl %[tmp_t2], %[tmp_t2] \n"
1702 "preceu.ph.qbr %[tmp_t7], %[tmp_t3] \n"
1703 "preceu.ph.qbl %[tmp_t3], %[tmp_t3] \n"
1704 "preceu.ph.qbr %[tmp_t8], %[tmp_t4] \n"
1705 "preceu.ph.qbl %[tmp_t4], %[tmp_t4] \n"
1706 "addu.ph %[tmp_t5], %[tmp_t5], %[tmp_t6] \n"
1707 "addu.ph %[tmp_t7], %[tmp_t7], %[tmp_t8] \n"
1708 "addu.ph %[tmp_t1], %[tmp_t1], %[tmp_t2] \n"
1709 "addu.ph %[tmp_t3], %[tmp_t3], %[tmp_t4] \n"
1710 "addu.ph %[tmp_t5], %[tmp_t5], %[tmp_t7] \n"
1711 "addu.ph %[tmp_t1], %[tmp_t1], %[tmp_t3] \n"
1712 "shrl.ph %[tmp_t5], %[tmp_t5], 2 \n"
1713 "shrl.ph %[tmp_t1], %[tmp_t1], 2 \n"
1714 "mult $ac0, %[const5], %[const5] \n"
1715 "mult $ac1, %[const5], %[const5] \n"
1716 "dpaq_s.w.ph $ac0, %[tmp_t5], %[const1] \n"
1717 "dpaq_s.w.ph $ac1, %[tmp_t5], %[const3] \n"
1718 "dpaq_s.w.ph $ac0, %[tmp_t1], %[const2] \n"
1719 "dpaq_s.w.ph $ac1, %[tmp_t1], %[const4] \n"
1720 "extr_r.w %[tmp_t7], $ac0, 9 \n"
1721 "extr_r.w %[tmp_t8], $ac1, 9 \n"
1722 "addiu %[src_rgb0], %[src_rgb0], 8 \n"
1723 "addiu %[src_rgb1], %[src_rgb1], 8 \n"
1724 "addiu %[dst_u], %[dst_u], 1 \n"
1725 "addiu %[dst_v], %[dst_v], 1 \n"
1726 "sb %[tmp_t7], -1(%[dst_u]) \n"
1727 "sb %[tmp_t8], -1(%[dst_v]) \n"
1728 ".set pop \n"
1729 :[tmp_t1] "=&r" (tmp_t1), [tmp_t2] "=&r" (tmp_t2),
1730 [tmp_t3] "=&r" (tmp_t3), [tmp_t4] "=&r" (tmp_t4),
1731 [tmp_t5] "=&r" (tmp_t5), [tmp_t6] "=&r" (tmp_t6),
1732 [tmp_t7] "=&r" (tmp_t7), [tmp_t8] "=&r" (tmp_t8),
1733 [src_rgb0] "+r" (src_rgb0), [src_rgb1] "+r" (src_rgb1),
1734 [dst_u] "+r" (dst_u), [dst_v] "+r" (dst_v)
1735 :[const1] "r" (const1), [const2] "r" (const2),
1736 [const3] "r" (const3), [const4] "r" (const4),
1737 [const5] "r" (const5)
1738 :"hi", "lo" ,"$ac1lo", "$ac1hi"
1739 );
1740 }
1741 }
1742
764 #endif // __mips_dsp_rev >= 2 1743 #endif // __mips_dsp_rev >= 2
765 1744
766 #endif // defined(__mips__) 1745 #endif // defined(__mips__)
767 1746
768 #ifdef __cplusplus 1747 #ifdef __cplusplus
769 } // extern "C" 1748 } // extern "C"
770 } // namespace libyuv 1749 } // namespace libyuv
771 #endif 1750 #endif
OLDNEW
« no previous file with comments | « source/row_any.cc ('k') | source/scale.cc » ('j') | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698