annotate ppc/vp3dsp_altivec.c @ 11032:01bd040f8607 libavcodec

Unroll main loop so the edge==0 case is seperate. This allows many things to be simplified away. h264 decoder is overall 1% faster with a mbaff sample and 0.1% slower with the cathedral sample, probably because the slow loop filter code must be loaded into the code cache for each first MB of each row but isnt used for the following MBs.

/*
 * Copyright (C) 2009 David Conrad
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
 * 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.
 *
 * FFmpeg 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
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with FFmpeg; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

#include "libavcodec/dsputil.h"
#include "util_altivec.h"
#include "types_altivec.h"

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)

void ff_vp3_idct_altivec(DCTELEM block[64])
{
    IDCT_START

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

    vec_st(b0, 0x00, block);
    vec_st(b1, 0x10, block);
    vec_st(b2, 0x20, block);
    vec_st(b3, 0x30, block);
    vec_st(b4, 0x40, block);
    vec_st(b5, 0x50, block);
    vec_st(b6, 0x60, block);
    vec_st(b7, 0x70, block);
}

void ff_vp3_idct_put_altivec(uint8_t *dst, int stride, DCTELEM block[64])
{
    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)
}

void ff_vp3_idct_add_altivec(uint8_t *dst, int stride, DCTELEM block[64])
{
    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)
}