1.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     1.2 +++ b/libavcodec/ppc/util_altivec.h	Tue Sep 25 15:55:33 2012 +0200
     1.3 @@ -0,0 +1,105 @@
     1.4 +/*
     1.5 + * This file is part of FFmpeg.
     1.6 + *
     1.7 + * FFmpeg is free software; you can redistribute it and/or
     1.8 + * modify it under the terms of the GNU Lesser General Public
     1.9 + * License as published by the Free Software Foundation; either
    1.10 + * version 2.1 of the License, or (at your option) any later version.
    1.11 + *
    1.12 + * FFmpeg is distributed in the hope that it will be useful,
    1.13 + * but WITHOUT ANY WARRANTY; without even the implied warranty of
    1.14 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
    1.15 + * Lesser General Public License for more details.
    1.16 + *
    1.17 + * You should have received a copy of the GNU Lesser General Public
    1.18 + * License along with FFmpeg; if not, write to the Free Software
    1.19 + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
    1.20 + */
    1.21 +
    1.22 +/**
    1.23 + * @file
    1.24 + * Contains misc utility macros and inline functions
    1.25 + */
    1.26 +
    1.27 +#ifndef AVCODEC_PPC_UTIL_ALTIVEC_H
    1.28 +#define AVCODEC_PPC_UTIL_ALTIVEC_H
    1.29 +
    1.30 +#include <stdint.h>
    1.31 +
    1.32 +#include "config.h"
    1.33 +
    1.34 +#if HAVE_ALTIVEC_H
    1.35 +#include <altivec.h>
    1.36 +#endif
    1.37 +
    1.38 +// used to build registers permutation vectors (vcprm)
    1.39 +// the 's' are for words in the _s_econd vector
    1.40 +#define WORD_0 0x00,0x01,0x02,0x03
    1.41 +#define WORD_1 0x04,0x05,0x06,0x07
    1.42 +#define WORD_2 0x08,0x09,0x0a,0x0b
    1.43 +#define WORD_3 0x0c,0x0d,0x0e,0x0f
    1.44 +#define WORD_s0 0x10,0x11,0x12,0x13
    1.45 +#define WORD_s1 0x14,0x15,0x16,0x17
    1.46 +#define WORD_s2 0x18,0x19,0x1a,0x1b
    1.47 +#define WORD_s3 0x1c,0x1d,0x1e,0x1f
    1.48 +
    1.49 +#define vcprm(a,b,c,d) (const vector unsigned char){WORD_ ## a, WORD_ ## b, WORD_ ## c, WORD_ ## d}
    1.50 +#define vcii(a,b,c,d) (const vector float){FLOAT_ ## a, FLOAT_ ## b, FLOAT_ ## c, FLOAT_ ## d}
    1.51 +
    1.52 +// vcprmle is used to keep the same index as in the SSE version.
    1.53 +// it's the same as vcprm, with the index inversed
    1.54 +// ('le' is Little Endian)
    1.55 +#define vcprmle(a,b,c,d) vcprm(d,c,b,a)
    1.56 +
    1.57 +// used to build inverse/identity vectors (vcii)
    1.58 +// n is _n_egative, p is _p_ositive
    1.59 +#define FLOAT_n -1.
    1.60 +#define FLOAT_p 1.
    1.61 +
    1.62 +
    1.63 +// Transpose 8x8 matrix of 16-bit elements (in-place)
    1.64 +#define TRANSPOSE8(a,b,c,d,e,f,g,h) \
    1.65 +do { \
    1.66 +    vector signed short A1, B1, C1, D1, E1, F1, G1, H1; \
    1.67 +    vector signed short A2, B2, C2, D2, E2, F2, G2, H2; \
    1.68 + \
    1.69 +    A1 = vec_mergeh (a, e); \
    1.70 +    B1 = vec_mergel (a, e); \
    1.71 +    C1 = vec_mergeh (b, f); \
    1.72 +    D1 = vec_mergel (b, f); \
    1.73 +    E1 = vec_mergeh (c, g); \
    1.74 +    F1 = vec_mergel (c, g); \
    1.75 +    G1 = vec_mergeh (d, h); \
    1.76 +    H1 = vec_mergel (d, h); \
    1.77 + \
    1.78 +    A2 = vec_mergeh (A1, E1); \
    1.79 +    B2 = vec_mergel (A1, E1); \
    1.80 +    C2 = vec_mergeh (B1, F1); \
    1.81 +    D2 = vec_mergel (B1, F1); \
    1.82 +    E2 = vec_mergeh (C1, G1); \
    1.83 +    F2 = vec_mergel (C1, G1); \
    1.84 +    G2 = vec_mergeh (D1, H1); \
    1.85 +    H2 = vec_mergel (D1, H1); \
    1.86 + \
    1.87 +    a = vec_mergeh (A2, E2); \
    1.88 +    b = vec_mergel (A2, E2); \
    1.89 +    c = vec_mergeh (B2, F2); \
    1.90 +    d = vec_mergel (B2, F2); \
    1.91 +    e = vec_mergeh (C2, G2); \
    1.92 +    f = vec_mergel (C2, G2); \
    1.93 +    g = vec_mergeh (D2, H2); \
    1.94 +    h = vec_mergel (D2, H2); \
    1.95 +} while (0)
    1.96 +
    1.97 +
    1.98 +/** \brief loads unaligned vector \a *src with offset \a offset
    1.99 +    and returns it */
   1.100 +static inline vector unsigned char unaligned_load(int offset, uint8_t *src)
   1.101 +{
   1.102 +    register vector unsigned char first = vec_ld(offset, src);
   1.103 +    register vector unsigned char second = vec_ld(offset+15, src);
   1.104 +    register vector unsigned char mask = vec_lvsl(offset, src);
   1.105 +    return vec_perm(first, second, mask);
   1.106 +}
   1.107 +
   1.108 +#endif /* AVCODEC_PPC_UTIL_ALTIVEC_H */