vp3dsp_altivec.c 5.28 KB
Newer Older
David Conrad's avatar
David Conrad committed
1 2 3
/*
 * Copyright (C) 2009 David Conrad
 *
4
 * This file is part of Libav.
David Conrad's avatar
David Conrad committed
5
 *
6
 * Libav is free software; you can redistribute it and/or
David Conrad's avatar
David Conrad committed
7 8 9 10
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
11
 * Libav is distributed in the hope that it will be useful,
David Conrad's avatar
David Conrad committed
12 13 14 15 16
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with Libav; if not, write to the Free Software
David Conrad's avatar
David Conrad committed
18 19 20
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

21 22 23 24 25 26 27
#include "config.h"
#include "libavutil/attributes.h"
#include "libavutil/cpu.h"
#include "libavcodec/vp3dsp.h"

#if HAVE_ALTIVEC

28 29
#include "libavutil/ppc/types_altivec.h"
#include "libavutil/ppc/util_altivec.h"
David Conrad's avatar
David Conrad committed
30
#include "libavcodec/dsputil.h"
31
#include "dsputil_altivec.h"
David Conrad's avatar
David Conrad committed
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

static const vec_s16 constants =
    {0, 64277, 60547, 54491, 46341, 36410, 25080, 12785};
static const vec_u8 interleave_high =
    {0, 1, 16, 17, 4, 5, 20, 21, 8, 9, 24, 25, 12, 13, 28, 29};

#define IDCT_START \
    vec_s16 A, B, C, D, Ad, Bd, Cd, Dd, E, F, G, H;\
    vec_s16 Ed, Gd, Add, Bdd, Fd, Hd;\
    vec_s16 eight = vec_splat_s16(8);\
    vec_u16 four = vec_splat_u16(4);\
\
    vec_s16 C1 = vec_splat(constants, 1);\
    vec_s16 C2 = vec_splat(constants, 2);\
    vec_s16 C3 = vec_splat(constants, 3);\
    vec_s16 C4 = vec_splat(constants, 4);\
    vec_s16 C5 = vec_splat(constants, 5);\
    vec_s16 C6 = vec_splat(constants, 6);\
    vec_s16 C7 = vec_splat(constants, 7);\
\
    vec_s16 b0 = vec_ld(0x00, block);\
    vec_s16 b1 = vec_ld(0x10, block);\
    vec_s16 b2 = vec_ld(0x20, block);\
    vec_s16 b3 = vec_ld(0x30, block);\
    vec_s16 b4 = vec_ld(0x40, block);\
    vec_s16 b5 = vec_ld(0x50, block);\
    vec_s16 b6 = vec_ld(0x60, block);\
    vec_s16 b7 = vec_ld(0x70, block);

// these functions do (a*C)>>16
// things are tricky because a is signed, but C unsigned.
// M15 is used if C fits in 15 bit unsigned (C6,C7)
// M16 is used if C requires 16 bits unsigned
static inline vec_s16 M15(vec_s16 a, vec_s16 C)
{
    return (vec_s16)vec_perm(vec_mule(a,C), vec_mulo(a,C), interleave_high);
}
static inline vec_s16 M16(vec_s16 a, vec_s16 C)
{
    return vec_add(a, M15(a, C));
}

#define IDCT_1D(ADD, SHIFT)\
    A = vec_add(M16(b1, C1), M15(b7, C7));\
    B = vec_sub(M15(b1, C7), M16(b7, C1));\
    C = vec_add(M16(b3, C3), M16(b5, C5));\
    D = vec_sub(M16(b5, C3), M16(b3, C5));\
\
    Ad = M16(vec_sub(A, C), C4);\
    Bd = M16(vec_sub(B, D), C4);\
\
    Cd = vec_add(A, C);\
    Dd = vec_add(B, D);\
\
    E = ADD(M16(vec_add(b0, b4), C4));\
    F = ADD(M16(vec_sub(b0, b4), C4));\
\
    G = vec_add(M16(b2, C2), M15(b6, C6));\
    H = vec_sub(M15(b2, C6), M16(b6, C2));\
\
    Ed = vec_sub(E, G);\
    Gd = vec_add(E, G);\
\
    Add = vec_add(F, Ad);\
    Bdd = vec_sub(Bd, H);\
\
    Fd = vec_sub(F, Ad);\
    Hd = vec_add(Bd, H);\
\
    b0 = SHIFT(vec_add(Gd, Cd));\
    b7 = SHIFT(vec_sub(Gd, Cd));\
\
    b1 = SHIFT(vec_add(Add, Hd));\
    b2 = SHIFT(vec_sub(Add, Hd));\
\
    b3 = SHIFT(vec_add(Ed, Dd));\
    b4 = SHIFT(vec_sub(Ed, Dd));\
\
    b5 = SHIFT(vec_add(Fd, Bdd));\
    b6 = SHIFT(vec_sub(Fd, Bdd));

#define NOP(a) a
#define ADD8(a) vec_add(a, eight)
#define SHIFT4(a) vec_sra(a, four)

117
static void vp3_idct_put_altivec(uint8_t *dst, int stride, DCTELEM block[64])
David Conrad's avatar
David Conrad committed
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
{
    vec_u8 t;
    IDCT_START

    // pixels are signed; so add 128*16 in addition to the normal 8
    vec_s16 v2048 = vec_sl(vec_splat_s16(1), vec_splat_u16(11));
    eight = vec_add(eight, v2048);

    IDCT_1D(NOP, NOP)
    TRANSPOSE8(b0, b1, b2, b3, b4, b5, b6, b7);
    IDCT_1D(ADD8, SHIFT4)

#define PUT(a)\
    t = vec_packsu(a, a);\
    vec_ste((vec_u32)t, 0, (unsigned int *)dst);\
    vec_ste((vec_u32)t, 4, (unsigned int *)dst);

    PUT(b0)     dst += stride;
    PUT(b1)     dst += stride;
    PUT(b2)     dst += stride;
    PUT(b3)     dst += stride;
    PUT(b4)     dst += stride;
    PUT(b5)     dst += stride;
    PUT(b6)     dst += stride;
    PUT(b7)
}

145
static void vp3_idct_add_altivec(uint8_t *dst, int stride, DCTELEM block[64])
David Conrad's avatar
David Conrad committed
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
{
    LOAD_ZERO;
    vec_u8 t, vdst;
    vec_s16 vdst_16;
    vec_u8 vdst_mask = vec_mergeh(vec_splat_u8(-1), vec_lvsl(0, dst));

    IDCT_START

    IDCT_1D(NOP, NOP)
    TRANSPOSE8(b0, b1, b2, b3, b4, b5, b6, b7);
    IDCT_1D(ADD8, SHIFT4)

#define ADD(a)\
    vdst = vec_ld(0, dst);\
    vdst_16 = (vec_s16)vec_perm(vdst, zero_u8v, vdst_mask);\
    vdst_16 = vec_adds(a, vdst_16);\
    t = vec_packsu(vdst_16, vdst_16);\
    vec_ste((vec_u32)t, 0, (unsigned int *)dst);\
    vec_ste((vec_u32)t, 4, (unsigned int *)dst);

    ADD(b0)     dst += stride;
    ADD(b1)     dst += stride;
    ADD(b2)     dst += stride;
    ADD(b3)     dst += stride;
    ADD(b4)     dst += stride;
    ADD(b5)     dst += stride;
    ADD(b6)     dst += stride;
    ADD(b7)
}
175 176 177 178 179

#endif /* HAVE_ALTIVEC */

av_cold void ff_vp3dsp_init_ppc(VP3DSPContext *c, int flags)
{
180 181
#if HAVE_ALTIVEC
    if (av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC) {
182 183 184 185
        c->idct_put  = vp3_idct_put_altivec;
        c->idct_add  = vp3_idct_add_altivec;
        c->idct_perm = FF_TRANSPOSE_IDCT_PERM;
    }
186
#endif
187
}