1.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     1.2 +++ b/libavcodec/arm/mathops.h	Tue Sep 25 15:55:33 2012 +0200
     1.3 @@ -0,0 +1,116 @@
     1.4 +/*
     1.5 + * simple math operations
     1.6 + * Copyright (c) 2006 Michael Niedermayer <michaelni@gmx.at> et al
     1.7 + *
     1.8 + * This file is part of FFmpeg.
     1.9 + *
    1.10 + * FFmpeg is free software; you can redistribute it and/or
    1.11 + * modify it under the terms of the GNU Lesser General Public
    1.12 + * License as published by the Free Software Foundation; either
    1.13 + * version 2.1 of the License, or (at your option) any later version.
    1.14 + *
    1.15 + * FFmpeg is distributed in the hope that it will be useful,
    1.16 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
    1.17 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
    1.18 + * Lesser General Public License for more details.
    1.19 + *
    1.20 + * You should have received a copy of the GNU Lesser General Public
    1.21 + * License along with FFmpeg; if not, write to the Free Software
    1.22 + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
    1.23 + */
    1.24 +
    1.25 +#ifndef AVCODEC_ARM_MATHOPS_H
    1.26 +#define AVCODEC_ARM_MATHOPS_H
    1.27 +
    1.28 +#include <stdint.h>
    1.29 +#include "config.h"
    1.30 +#include "libavutil/common.h"
    1.31 +
    1.32 +#if HAVE_INLINE_ASM
    1.33 +
    1.34 +#   define MULL MULL
    1.35 +static inline av_const int MULL(int a, int b, unsigned shift)
    1.36 +{
    1.37 +    int lo, hi;
    1.38 +    __asm__("smull %0, %1, %2, %3     \n\t"
    1.39 +            "mov   %0, %0,     lsr %4 \n\t"
    1.40 +            "add   %1, %0, %1, lsl %5 \n\t"
    1.41 +            : "=&r"(lo), "=&r"(hi)
    1.42 +            : "r"(b), "r"(a), "ir"(shift), "ir"(32-shift));
    1.43 +    return hi;
    1.44 +}
    1.45 +
    1.46 +#define MULH MULH
    1.47 +#if HAVE_ARMV6
    1.48 +static inline av_const int MULH(int a, int b)
    1.49 +{
    1.50 +    int r;
    1.51 +    __asm__ ("smmul %0, %1, %2" : "=r"(r) : "r"(a), "r"(b));
    1.52 +    return r;
    1.53 +}
    1.54 +#else
    1.55 +static inline av_const int MULH(int a, int b)
    1.56 +{
    1.57 +    int lo, hi;
    1.58 +    __asm__ ("smull %0, %1, %2, %3" : "=&r"(lo), "=&r"(hi) : "r"(b), "r"(a));
    1.59 +    return hi;
    1.60 +}
    1.61 +#endif
    1.62 +
    1.63 +static inline av_const int64_t MUL64(int a, int b)
    1.64 +{
    1.65 +    union { uint64_t x; unsigned hl[2]; } x;
    1.66 +    __asm__ ("smull %0, %1, %2, %3"
    1.67 +             : "=r"(x.hl[0]), "=r"(x.hl[1]) : "r"(a), "r"(b));
    1.68 +    return x.x;
    1.69 +}
    1.70 +#define MUL64 MUL64
    1.71 +
    1.72 +static inline av_const int64_t MAC64(int64_t d, int a, int b)
    1.73 +{
    1.74 +    union { uint64_t x; unsigned hl[2]; } x = { d };
    1.75 +    __asm__ ("smlal %0, %1, %2, %3"
    1.76 +             : "+r"(x.hl[0]), "+r"(x.hl[1]) : "r"(a), "r"(b));
    1.77 +    return x.x;
    1.78 +}
    1.79 +#define MAC64(d, a, b) ((d) = MAC64(d, a, b))
    1.80 +#define MLS64(d, a, b) MAC64(d, -(a), b)
    1.81 +
    1.82 +#if HAVE_ARMV5TE
    1.83 +
    1.84 +/* signed 16x16 -> 32 multiply add accumulate */
    1.85 +#   define MAC16(rt, ra, rb)                                            \
    1.86 +    __asm__ ("smlabb %0, %1, %2, %0" : "+r"(rt) : "r"(ra), "r"(rb));
    1.87 +
    1.88 +/* signed 16x16 -> 32 multiply */
    1.89 +#   define MUL16 MUL16
    1.90 +static inline av_const int MUL16(int ra, int rb)
    1.91 +{
    1.92 +    int rt;
    1.93 +    __asm__ ("smulbb %0, %1, %2" : "=r"(rt) : "r"(ra), "r"(rb));
    1.94 +    return rt;
    1.95 +}
    1.96 +
    1.97 +#endif
    1.98 +
    1.99 +#define mid_pred mid_pred
   1.100 +static inline av_const int mid_pred(int a, int b, int c)
   1.101 +{
   1.102 +    int m;
   1.103 +    __asm__ volatile (
   1.104 +        "mov   %0, %2  \n\t"
   1.105 +        "cmp   %1, %2  \n\t"
   1.106 +        "movgt %0, %1  \n\t"
   1.107 +        "movgt %1, %2  \n\t"
   1.108 +        "cmp   %1, %3  \n\t"
   1.109 +        "movle %1, %3  \n\t"
   1.110 +        "cmp   %0, %1  \n\t"
   1.111 +        "movgt %0, %1  \n\t"
   1.112 +        : "=&r"(m), "+r"(a)
   1.113 +        : "r"(b), "r"(c));
   1.114 +    return m;
   1.115 +}
   1.116 +
   1.117 +#endif /* HAVE_INLINE_ASM */
   1.118 +
   1.119 +#endif /* AVCODEC_ARM_MATHOPS_H */