view dct.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.

/*
 * (I)DCT Transforms
 * Copyright (c) 2009 Peter Ross <[e-mail]>
 * Copyright (c) 2010 Alex Converse <[e-mail]>
 * Copyright (c) 2010 Vitor Sessak
 *
 * 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 St, Fifth Floor, Boston, MA  02110-1301  USA
 */

/**
 * @file libavcodec/dct.c
 * (Inverse) Discrete Cosine Transforms. These are also known as the
 * type II and type III DCTs respectively.
 */

#include <math.h>
#include "dsputil.h"

av_cold int ff_dct_init(DCTContext *s, int nbits, int inverse)
{
    int n = 1 << nbits;
    int i;

    s->nbits    = nbits;
    s->inverse  = inverse;

    ff_init_ff_cos_tabs(nbits+2);

    s->costab = ff_cos_tabs[nbits+2];

    s->csc2 = av_malloc(n/2 * sizeof(FFTSample));

    if (ff_rdft_init(&s->rdft, nbits, inverse) < 0) {
        av_free(s->csc2);
        return -1;
    }

    for (i = 0; i < n/2; i++)
        s->csc2[i] = 0.5 / sin((M_PI / (2*n) * (2*i + 1)));

    return 0;
}

/* sin((M_PI * x / (2*n)) */
#define SIN(s,n,x) (s->costab[(n) - (x)])

/* cos((M_PI * x / (2*n)) */
#define COS(s,n,x) (s->costab[x])

static void ff_dct_calc_c(DCTContext *ctx, FFTSample *data)
{
    int n = 1 << ctx->nbits;
    int i;

    if (ctx->inverse) {
        float next = data[n - 1];
        float inv_n = 1.0f / n;

        for (i = n - 2; i >= 2; i -= 2) {
            float val1 = data[i    ];
            float val2 = data[i - 1] - data[i + 1];
            float c = COS(ctx, n, i);
            float s = SIN(ctx, n, i);

            data[i    ] = c * val1 + s * val2;
            data[i + 1] = s * val1 - c * val2;
        }

        data[1] = 2 * next;

        ff_rdft_calc(&ctx->rdft, data);

        for (i = 0; i < n / 2; i++) {
            float tmp1 = data[i        ] * inv_n;
            float tmp2 = data[n - i - 1] * inv_n;
            float csc = ctx->csc2[i] * (tmp1 - tmp2);

            tmp1 += tmp2;
            data[i        ] = tmp1 + csc;
            data[n - i - 1] = tmp1 - csc;
        }
    } else {
        float next;
        for (i=0; i < n/2; i++) {
            float tmp1 = data[i        ];
            float tmp2 = data[n - i - 1];
            float s = SIN(ctx, n, 2*i + 1);

            s *= tmp1 - tmp2;
            tmp1 = (tmp1 + tmp2) * 0.5f;

            data[i    ] = tmp1 + s;
            data[n-i-1] = tmp1 - s;
        }

        ff_rdft_calc(&ctx->rdft, data);

        next = data[1] * 0.5;
        data[1] *= -1;

        for (i = n - 2; i >= 0; i -= 2) {
            float inr = data[i    ];
            float ini = data[i + 1];
            float c = COS(ctx, n, i);
            float s = SIN(ctx, n, i);

            data[i  ] = c * inr + s * ini;

            data[i+1] = next;

            next +=     s * inr - c * ini;
        }
    }
}

void ff_dct_calc(DCTContext *s, FFTSample *data)
{
    ff_dct_calc_c(s, data);
}

av_cold void ff_dct_end(DCTContext *s)
{
    ff_rdft_end(&s->rdft);
    av_free(s->csc2);
}