rv34.c

ffmpeg移植到symbian的全部源代码

/*
 * RV30/40 decoder common data
 * Copyright (c) 2007 Mike Melanson, Konstantin Shishkov
 *
 * 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
 */

/**
 * @file rv34.c
 * RV30/40 decoder common data
 */

#include "avcodec.h"
#include "dsputil.h"
#include "mpegvideo.h"
#include "golomb.h"
#include "rectangle.h"

#include "rv34vlc.h"
#include "rv34data.h"
#include "rv34.h"

//#define DEBUG

/** translation of RV30/40 macroblock types to lavc ones */
static const int rv34_mb_type_to_lavc[12] = {
    MB_TYPE_INTRA,
    MB_TYPE_INTRA16x16,
    MB_TYPE_16x16   | MB_TYPE_L0,
    MB_TYPE_8x8     | MB_TYPE_L0,
    MB_TYPE_16x16   | MB_TYPE_L0,
    MB_TYPE_16x16   | MB_TYPE_L1,
    MB_TYPE_SKIP,
    MB_TYPE_DIRECT2 | MB_TYPE_16x16,
    MB_TYPE_16x8    | MB_TYPE_L0,
    MB_TYPE_8x16    | MB_TYPE_L0,
    MB_TYPE_16x16   | MB_TYPE_L0L1,
    MB_TYPE_16x16   | MB_TYPE_L0
};


static RV34VLC intra_vlcs[NUM_INTRA_TABLES], inter_vlcs[NUM_INTER_TABLES];

/**
 * @defgroup vlc RV30/40 VLC generating functions
 * @{
 */

/**
 * Generate VLC from codeword lengths.
 * @param bits   codeword lengths (zeroes are accepted)
 * @param size   length of input data
 * @param insyms symbols for input codes (NULL for default ones)
 */
static void rv34_gen_vlc(const uint8_t *bits, int size, VLC *vlc, const uint8_t *insyms)
{
    int i;
    int counts[17] = {0}, codes[17];
#ifdef __CW32__
    uint16_t *cw, *syms;
    uint8_t *bits2;
#else
    uint16_t cw[size], syms[size];
    uint8_t bits2[size];
#endif
    int maxbits = 0, realsize = 0;
#ifdef __CW32__
    cw = av_malloc(sizeof(uint16_t)*size);
    syms = av_malloc(sizeof(uint16_t)*size);
    bits2 = av_malloc(sizeof(uint8_t)*size);
#endif

    for(i = 0; i < size; i++){
        if(bits[i]){
            bits2[realsize] = bits[i];
            syms[realsize] = insyms ? insyms[i] : i;
            realsize++;
            maxbits = FFMAX(maxbits, bits[i]);
            counts[bits[i]]++;
        }
    }

    codes[0] = 0;
    for(i = 0; i < 16; i++)
        codes[i+1] = (codes[i] + counts[i]) << 1;
    for(i = 0; i < realsize; i++)
        cw[i] = codes[bits2[i]]++;

    init_vlc_sparse(vlc, FFMIN(maxbits, 9), realsize,
                    bits2, 1, 1,
                    cw,    2, 2,
                    syms,  2, 2, INIT_VLC_USE_STATIC);
#ifdef __CW32__
    av_free(cw);
    av_free(syms);
    av_free(bits2);
#endif
}

/**
 * Initialize all tables.
 */
static av_cold void rv34_init_tables()
{
    int i, j, k;

    for(i = 0; i < NUM_INTRA_TABLES; i++){
        for(j = 0; j < 2; j++){
            rv34_gen_vlc(rv34_table_intra_cbppat   [i][j], CBPPAT_VLC_SIZE,   &intra_vlcs[i].cbppattern[j],     NULL);
            rv34_gen_vlc(rv34_table_intra_secondpat[i][j], OTHERBLK_VLC_SIZE, &intra_vlcs[i].second_pattern[j], NULL);
            rv34_gen_vlc(rv34_table_intra_thirdpat [i][j], OTHERBLK_VLC_SIZE, &intra_vlcs[i].third_pattern[j],  NULL);
            for(k = 0; k < 4; k++)
                rv34_gen_vlc(rv34_table_intra_cbp[i][j+k*2],  CBP_VLC_SIZE,   &intra_vlcs[i].cbp[j][k],         rv34_cbp_code);
        }
        for(j = 0; j < 4; j++)
            rv34_gen_vlc(rv34_table_intra_firstpat[i][j], FIRSTBLK_VLC_SIZE, &intra_vlcs[i].first_pattern[j], NULL);
        rv34_gen_vlc(rv34_intra_coeff[i], COEFF_VLC_SIZE, &intra_vlcs[i].coefficient, NULL);
    }

    for(i = 0; i < NUM_INTER_TABLES; i++){
        rv34_gen_vlc(rv34_inter_cbppat[i], CBPPAT_VLC_SIZE, &inter_vlcs[i].cbppattern[0], NULL);
        for(j = 0; j < 4; j++)
            rv34_gen_vlc(rv34_inter_cbp[i][j], CBP_VLC_SIZE, &inter_vlcs[i].cbp[0][j], rv34_cbp_code);
        for(j = 0; j < 2; j++){
            rv34_gen_vlc(rv34_table_inter_firstpat [i][j], FIRSTBLK_VLC_SIZE, &inter_vlcs[i].first_pattern[j],  NULL);
            rv34_gen_vlc(rv34_table_inter_secondpat[i][j], OTHERBLK_VLC_SIZE, &inter_vlcs[i].second_pattern[j], NULL);
            rv34_gen_vlc(rv34_table_inter_thirdpat [i][j], OTHERBLK_VLC_SIZE, &inter_vlcs[i].third_pattern[j],  NULL);
        }
        rv34_gen_vlc(rv34_inter_coeff[i], COEFF_VLC_SIZE, &inter_vlcs[i].coefficient, NULL);
    }
}

/** @} */ // vlc group


/**
 * @defgroup transform RV30/40 inverse transform functions
 * @{
 */

static av_always_inline void rv34_row_transform(int temp[16], DCTELEM *block)
{
    int i;

    for(i=0; i<4; i++){
        const int z0= 13*(block[i+8*0] +    block[i+8*2]);
        const int z1= 13*(block[i+8*0] -    block[i+8*2]);
        const int z2=  7* block[i+8*1] - 17*block[i+8*3];
        const int z3= 17* block[i+8*1] +  7*block[i+8*3];

        temp[4*i+0]= z0+z3;
        temp[4*i+1]= z1+z2;
        temp[4*i+2]= z1-z2;
        temp[4*i+3]= z0-z3;
    }
}

/**
 * Real Video 3.0/4.0 inverse transform
 * Code is almost the same as in SVQ3, only scaling is different.
 */
static void rv34_inv_transform(DCTELEM *block){
    int temp[16];
    int i;

    rv34_row_transform(temp, block);

    for(i=0; i<4; i++){
        const int z0= 13*(temp[4*0+i] +    temp[4*2+i]) + 0x200;
        const int z1= 13*(temp[4*0+i] -    temp[4*2+i]) + 0x200;
        const int z2=  7* temp[4*1+i] - 17*temp[4*3+i];
        const int z3= 17* temp[4*1+i] +  7*temp[4*3+i];

        block[i*8+0]= (z0 + z3)>>10;
        block[i*8+1]= (z1 + z2)>>10;
        block[i*8+2]= (z1 - z2)>>10;
        block[i*8+3]= (z0 - z3)>>10;
    }

}

/**
 * RealVideo 3.0/4.0 inverse transform for DC block
 *
 * Code is almost the same as rv34_inv_transform()
 * but final coefficients are multiplied by 1.5 and have no rounding.
 */
static void rv34_inv_transform_noround(DCTELEM *block){
    int temp[16];
    int i;

    rv34_row_transform(temp, block);

    for(i=0; i<4; i++){
        const int z0= 13*(temp[4*0+i] +    temp[4*2+i]);
        const int z1= 13*(temp[4*0+i] -    temp[4*2+i]);
        const int z2=  7* temp[4*1+i] - 17*temp[4*3+i];
        const int z3= 17* temp[4*1+i] +  7*temp[4*3+i];

        block[i*8+0]= ((z0 + z3)*3)>>11;
        block[i*8+1]= ((z1 + z2)*3)>>11;
        block[i*8+2]= ((z1 - z2)*3)>>11;
        block[i*8+3]= ((z0 - z3)*3)>>11;
    }

}

/** @} */ // transform


/**
 * @defgroup block RV30/40 4x4 block decoding functions
 * @{
 */

/**
 * Decode coded block pattern.
 */
static int rv34_decode_cbp(GetBitContext *gb, RV34VLC *vlc, int table)
{
    int pattern, code, cbp=0;
    int ones;
    static const int cbp_masks[3] = {0x100000, 0x010000, 0x110000};
    static const int shifts[4] = { 0, 2, 8, 10 };
#ifdef __CW32__
    int *curshift = (int*)shifts;
#else
    int *curshift = shifts;
#endif
    int i, t, mask;

    code = get_vlc2(gb, vlc->cbppattern[table].table, 9, 2);
    pattern = code & 0xF;
    code >>= 4;

    ones = rv34_count_ones[pattern];

    for(mask = 8; mask; mask >>= 1, curshift++){
        if(pattern & mask)
            cbp |= get_vlc2(gb, vlc->cbp[table][ones].table, vlc->cbp[table][ones].bits, 1) << curshift[0];
    }

    for(i = 0; i < 4; i++){
        t = modulo_three_table[code][i];
        if(t == 1)
            cbp |= cbp_masks[get_bits1(gb)] << i;
        if(t == 2)
            cbp |= cbp_masks[2] << i;
    }
    return cbp;
}

/**
 * Get one coefficient value from the bistream and store it.
 */
static inline void decode_coeff(DCTELEM *dst, int coef, int esc, GetBitContext *gb, VLC* vlc)
{
    if(coef){
        if(coef == esc){
            coef = get_vlc2(gb, vlc->table, 9, 2);
            if(coef > 23){
                coef -= 23;
                coef = 22 + ((1 << coef) | get_bits(gb, coef));
            }
            coef += esc;
        }
        if(get_bits1(gb))
            coef = -coef;
        *dst = coef;
    }
}

/**
 * Decode 2x2 subblock of coefficients.
 */
static inline void decode_subblock(DCTELEM *dst, int code, const int is_block2, GetBitContext *gb, VLC *vlc)
{
    int coeffs[4];

    coeffs[0] = modulo_three_table[code][0];
    coeffs[1] = modulo_three_table[code][1];
    coeffs[2] = modulo_three_table[code][2];
    coeffs[3] = modulo_three_table[code][3];
    decode_coeff(dst  , coeffs[0], 3, gb, vlc);
    if(is_block2){
        decode_coeff(dst+8, coeffs[1], 2, gb, vlc);
        decode_coeff(dst+1, coeffs[2], 2, gb, vlc);
    }else{
        decode_coeff(dst+1, coeffs[1], 2, gb, vlc);
        decode_coeff(dst+8, coeffs[2], 2, gb, vlc);
    }
    decode_coeff(dst+9, coeffs[3], 2, gb, vlc);
}

/**
 * Decode coefficients for 4x4 block.
 *
 * This is done by filling 2x2 subblocks with decoded coefficients
 * in this order (the same for subblocks and subblock coefficients):
 *  o--o
 *    /
 *   /
 *  o--o
 */

static inline void rv34_decode_block(DCTELEM *dst, GetBitContext *gb, RV34VLC *rvlc, int fc, int sc)
{
    int code, pattern;

    code = get_vlc2(gb, rvlc->first_pattern[fc].table, 9, 2);

    pattern = code & 0x7;

    code >>= 3;
    decode_subblock(dst, code, 0, gb, &rvlc->coefficient);

    if(pattern & 4){
        code = get_vlc2(gb, rvlc->second_pattern[sc].table, 9, 2);
        decode_subblock(dst + 2, code, 0, gb, &rvlc->coefficient);
    }
    if(pattern & 2){ // Looks like coefficients 1 and 2 are swapped for this block
        code = get_vlc2(gb, rvlc->second_pattern[sc].table, 9, 2);
        decode_subblock(dst + 8*2, code, 1, gb, &rvlc->coefficient);
    }
    if(pattern & 1){