motion_est_mvi_asm.S 6.17 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186
/*
 * Alpha optimized DSP utils
 * Copyright (c) 2002 Falk Hueffner <falk@debian.org>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */

#include "regdef.h"
#ifdef HAVE_AV_CONFIG_H	
#include "config.h"
#endif

/* Some nicer register names.  */
#define ta t10
#define tb t11
#define tc t12
#define td AT
/* Danger: these overlap with the argument list and the return value */
#define te a5
#define tf a4
#define tg a3
#define th v0
        
        .set noat
        .set noreorder
        .arch pca56
        .text

/*****************************************************************************
 * int pix_abs16x16_mvi_asm(uint8_t *pix1, uint8_t *pix2, int line_size)
 *
 * This code is written with a pca56 in mind. For ev6, one should
 * really take the increased latency of 3 cycles for MVI instructions
 * into account.
 *
 * It is important to keep the loading and first use of a register as
 * far apart as possible, because if a register is accessed before it
 * has been fetched from memory, the CPU will stall.
 */
        .align 4
        .globl pix_abs16x16_mvi_asm
        .ent pix_abs16x16_mvi_asm
pix_abs16x16_mvi_asm:
        .frame sp, 0, ra, 0
        .prologue 0

#ifdef HAVE_GPROF
        lda     AT, _mcount
        jsr     AT, (AT), _mcount
#endif

        and     a1, 7, t0
        clr     v0
        lda     a3, 16
        beq     t0, $aligned
        .align 4
$unaligned:
        /* Registers:
           line 0:
           t0:  left_u -> left lo -> left
           t1:  mid
           t2:  right_u -> right hi -> right
           t3:  ref left
           t4:  ref right
           line 1:
           t5:  left_u -> left lo -> left
           t6:  mid
           t7:  right_u -> right hi -> right
           t8:  ref left
           t9:  ref right
           temp:
           ta:  left hi
           tb:  right lo
           tc:  error left
           td:  error right  */

        /* load line 0 */
        ldq_u   t0, 0(a1)       # left_u
        ldq_u   t1, 8(a1)       # mid
        ldq_u   t2, 16(a1)      # right_u
        ldq     t3, 0(a0)       # ref left
        ldq     t4, 8(a0)       # ref right
        addq    a0, a2, a0      # pix1
        addq    a1, a2, a1      # pix2
        /* load line 1 */        
        ldq_u   t5, 0(a1)       # left_u
        ldq_u   t6, 8(a1)       # mid
        ldq_u   t7, 16(a1)      # right_u
        ldq     t8, 0(a0)       # ref left
        ldq     t9, 8(a0)       # ref right
        addq    a0, a2, a0      # pix1
        addq    a1, a2, a1      # pix2
        /* calc line 0 */
        extql   t0, a1, t0      # left lo
        extqh   t1, a1, ta      # left hi
        extql   t1, a1, tb      # right lo
        or      t0, ta, t0      # left
        extqh   t2, a1, t2      # right hi
        perr    t3, t0, tc      # error left
        or      t2, tb, t2      # right
        perr    t4, t2, td      # error right
        addq    v0, tc, v0      # add error left
        addq    v0, td, v0      # add error left
        /* calc line 1 */
        extql   t5, a1, t5      # left lo
        extqh   t6, a1, ta      # left hi
        extql   t6, a1, tb      # right lo
        or      t5, ta, t5      # left
        extqh   t7, a1, t7      # right hi
        perr    t8, t5, tc      # error left
        or      t7, tb, t7      # right
        perr    t9, t7, td      # error right
        addq    v0, tc, v0      # add error left
        addq    v0, td, v0      # add error left
        /* loop */
        subq    a3,  2, a3      # h -= 2
        bne     a3, $unaligned
        ret

        .align 4
$aligned:
        /* load line 0 */
        ldq     t0, 0(a1)       # left
        ldq     t1, 8(a1)       # right
        addq    a1, a2, a1      # pix2
        ldq     t2, 0(a0)       # ref left
        ldq     t3, 8(a0)       # ref right
        addq    a0, a2, a0      # pix1
        /* load line 1 */
        ldq     t4, 0(a1)       # left
        ldq     t5, 8(a1)       # right
        addq    a1, a2, a1      # pix2
        ldq     t6, 0(a0)       # ref left
        ldq     t7, 8(a0)       # ref right
        addq    a0, a2, a0      # pix1
        /* load line 2 */
        ldq     t8, 0(a1)       # left
        ldq     t9, 8(a1)       # right
        addq    a1, a2, a1      # pix2
        ldq     ta, 0(a0)       # ref left
        ldq     tb, 8(a0)       # ref right
        addq    a0, a2, a0      # pix1
        /* load line 3 */
        ldq     tc, 0(a1)       # left
        ldq     td, 8(a1)       # right
        addq    a1, a2, a1      # pix2
        ldq     te, 0(a0)       # ref left
        ldq     tf, 8(a0)       # ref right
        /* calc line 0 */
        perr    t0, t2, t0      # error left
        addq    a0, a2, a0      # pix1
        perr    t1, t3, t1      # error right
        addq    v0, t0, v0      # add error left
        /* calc line 1 */
        perr    t4, t6, t0      # error left
        addq    v0, t1, v0      # add error right
        perr    t5, t7, t1      # error right
        addq    v0, t0, v0      # add error left
        /* calc line 2 */
        perr    t8, ta, t0      # error left
        addq    v0, t1, v0      # add error right
        perr    t9, tb, t1      # error right
        addq    v0, t0, v0      # add error left
        /* calc line 3 */
        perr    tc, te, t0      # error left
        addq    v0, t1, v0      # add error right
        perr    td, tf, t1      # error right
        addq    v0, t0, v0      # add error left
        addq    v0, t1, v0      # add error right
        /* loop */
        subq    a3,  4, a3      # h -= 4
        bne     a3, $aligned
        ret
        .end pix_abs16x16_mvi_asm