analyse.c

linux下编译已经通过

/*****************************************************************************
 * analyse.c: h264 encoder library
 *****************************************************************************
 * Copyright (C) 2003 x264 project
 * $Id: analyse.c,v 1.1 2004/06/03 19:27:08 fenrir Exp $
 *
 * Authors: Laurent Aimar <[e-mail]>
 *          Loren Merritt <[e-mail]>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program 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 General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111, USA.
 *****************************************************************************/

#include <math.h>
#include <limits.h>
#ifndef _MSC_VER
#include <unistd.h>
#endif

#include "common/common.h"
#include "macroblock.h"
#include "me.h"
#include "ratecontrol.h"
#include "analyse.h"
#include "rdo.c"

typedef struct
{
    /* 16x16 */
    int i_ref;
    int       i_rd16x16;
    x264_me_t me16x16;

    /* 8x8 */
    int       i_cost8x8;
    int       mvc[32][5][2]; /* [ref][0] is 16x16 mv,
                                [ref][1..4] are 8x8 mv from partition [0..3] */
    x264_me_t me8x8[4];

    /* Sub 4x4 */
    int       i_cost4x4[4]; /* cost per 8x8 partition */
    x264_me_t me4x4[4][4];

    /* Sub 8x4 */
    int       i_cost8x4[4]; /* cost per 8x8 partition */
    x264_me_t me8x4[4][2];

    /* Sub 4x8 */
    int       i_cost4x8[4]; /* cost per 8x8 partition */
    x264_me_t me4x8[4][2];

    /* 16x8 */
    int       i_cost16x8;
    x264_me_t me16x8[2];

    /* 8x16 */
    int       i_cost8x16;
    x264_me_t me8x16[2];

} x264_mb_analysis_list_t;

typedef struct
{
    /* conduct the analysis using this lamda and QP */
    int i_lambda;
    int i_lambda2;
    int i_qp;
    int16_t *p_cost_mv;
    int b_mbrd;


    /* I: Intra part */
    /* Take some shortcuts in intra search if intra is deemed unlikely */
    int b_fast_intra;
    int b_try_pskip;

    /* Luma part */
    int i_satd_i16x16;
    int i_satd_i16x16_dir[7];
    int i_predict16x16;

    int i_satd_i8x8;
    int i_satd_i8x8_dir[12][4];
    int i_predict8x8[4];

    int i_satd_i4x4;
    int i_predict4x4[16];

    /* Chroma part */
    int i_satd_i8x8chroma;
    int i_satd_i8x8chroma_dir[4];
    int i_predict8x8chroma;

    /* II: Inter part P/B frame */
    x264_mb_analysis_list_t l0;
    x264_mb_analysis_list_t l1;

    int i_cost16x16bi; /* used the same ref and mv as l0 and l1 (at least for now) */
    int i_cost16x16direct;
    int i_cost8x8bi;
    int i_cost8x8direct[4];
    int i_cost16x8bi;
    int i_cost8x16bi;
    int i_rd16x16bi;
    int i_rd16x16direct;
    int i_rd16x8bi;
    int i_rd8x16bi;
    int i_rd8x8bi;

    int i_mb_partition16x8[2]; /* mb_partition_e */
    int i_mb_partition8x16[2];
    int i_mb_type16x8; /* mb_class_e */
    int i_mb_type8x16;

    int b_direct_available;

} x264_mb_analysis_t;

/* lambda = pow(2,qp/6-2) */
static const int i_qp0_cost_table[52] = {
   1, 1, 1, 1, 1, 1, 1, 1,  /*  0-7 */
   1, 1, 1, 1,              /*  8-11 */
   1, 1, 1, 1, 2, 2, 2, 2,  /* 12-19 */
   3, 3, 3, 4, 4, 4, 5, 6,  /* 20-27 */
   6, 7, 8, 9,10,11,13,14,  /* 28-35 */
  16,18,20,23,25,29,32,36,  /* 36-43 */
  40,45,51,57,64,72,81,91   /* 44-51 */
};

/* pow(lambda,2) * .9 */
static const int i_qp0_cost2_table[52] = {
   1,   1,   1,   1,   1,   1, /*  0-5  */
   1,   1,   1,   1,   1,   1, /*  6-11 */
   1,   1,   1,   2,   2,   3, /* 12-17 */
   4,   5,   6,   7,   9,  11, /* 18-23 */
  14,  18,  23,  29,  36,  46, /* 24-29 */
  58,  73,  91, 115, 145, 183, /* 30-35 */
 230, 290, 366, 461, 581, 731, /* 36-41 */
 922,1161,1463,1843,2322,2926, /* 42-47 */
3686,4645,5852,7373
};

/* TODO: calculate CABAC costs */
static const int i_mb_b_cost_table[X264_MBTYPE_MAX] = {
    9, 9, 9, 9, 0, 0, 0, 1, 3, 7, 7, 7, 3, 7, 7, 7, 5, 9, 0
};
static const int i_mb_b16x8_cost_table[17] = {
    0, 0, 0, 0, 0, 0, 0, 0, 5, 7, 7, 7, 5, 7, 9, 9, 9
};
static const int i_sub_mb_b_cost_table[13] = {
    7, 5, 5, 3, 7, 5, 7, 3, 7, 7, 7, 5, 1
};
static const int i_sub_mb_p_cost_table[4] = {
    5, 3, 3, 1
};

static void x264_analyse_update_cache( x264_t *h, x264_mb_analysis_t *a );

/* initialize an array of lambda*nbits for all possible mvs */
static void x264_mb_analyse_load_costs( x264_t *h, x264_mb_analysis_t *a )
{
    static int16_t *p_cost_mv[52];

    if( !p_cost_mv[a->i_qp] )
    {
        /* could be faster, but isn't called many times */
        /* factor of 4 from qpel, 2 from sign, and 2 because mv can be opposite from mvp */
        int i;
        p_cost_mv[a->i_qp] = x264_malloc( (4*4*2048 + 1) * sizeof(int16_t) );
        p_cost_mv[a->i_qp] += 2*4*2048;
        for( i = 0; i <= 2*4*2048; i++ )
        {
            p_cost_mv[a->i_qp][-i] =
            p_cost_mv[a->i_qp][i]  = a->i_lambda * bs_size_se( i );
        }
    }

    a->p_cost_mv = p_cost_mv[a->i_qp];
}

static void x264_mb_analyse_init( x264_t *h, x264_mb_analysis_t *a, int i_qp )
{
    memset( a, 0, sizeof( x264_mb_analysis_t ) );

    /* conduct the analysis using this lamda and QP */
    a->i_qp = h->mb.i_qp = i_qp;
    h->mb.i_chroma_qp = i_chroma_qp_table[x264_clip3( i_qp + h->pps->i_chroma_qp_index_offset, 0, 51 )];
    a->i_lambda = i_qp0_cost_table[i_qp];
    a->i_lambda2 = i_qp0_cost2_table[i_qp];
    a->b_mbrd = h->param.analyse.i_subpel_refine >= 6 &&
                ( h->sh.i_type != SLICE_TYPE_B || h->param.analyse.b_bframe_rdo );

    h->mb.i_me_method = h->param.analyse.i_me_method;
    h->mb.i_subpel_refine = h->param.analyse.i_subpel_refine;
    h->mb.b_chroma_me = h->param.analyse.b_chroma_me && h->sh.i_type == SLICE_TYPE_P
                        && h->mb.i_subpel_refine >= 5;
    h->mb.b_trellis = h->param.analyse.i_trellis > 1 && a->b_mbrd;
    h->mb.b_transform_8x8 = 0;
    h->mb.b_noise_reduction = 0;

    /* I: Intra part */
    a->i_satd_i16x16 =
    a->i_satd_i8x8   =
    a->i_satd_i4x4   =
    a->i_satd_i8x8chroma = COST_MAX;

    a->b_fast_intra = 0;

    /* II: Inter part P/B frame */
    if( h->sh.i_type != SLICE_TYPE_I )
    {
        int i, j;
        int i_fmv_range = 4 * h->param.analyse.i_mv_range;
        // limit motion search to a slightly smaller range than the theoretical limit,
        // since the search may go a few iterations past its given range
        int i_fpel_border = 5; // umh unconditional radius
        int i_spel_border = 8; // 1.5 for subpel_satd, 1.5 for subpel_rd, 2 for bime, round up

        /* Calculate max allowed MV range */
#define CLIP_FMV(mv) x264_clip3( mv, -i_fmv_range, i_fmv_range )
        h->mb.mv_min[0] = 4*( -16*h->mb.i_mb_x - 24 );
        h->mb.mv_max[0] = 4*( 16*( h->sps->i_mb_width - h->mb.i_mb_x - 1 ) + 24 );
        h->mb.mv_min_spel[0] = CLIP_FMV( h->mb.mv_min[0] );
        h->mb.mv_max_spel[0] = CLIP_FMV( h->mb.mv_max[0] );
        h->mb.mv_min_fpel[0] = (h->mb.mv_min_spel[0]>>2) + i_fpel_border;
        h->mb.mv_max_fpel[0] = (h->mb.mv_max_spel[0]>>2) - i_fpel_border;
        if( h->mb.i_mb_x == 0)
        {
            int mb_y = h->mb.i_mb_y >> h->sh.b_mbaff;
            int mb_height = h->sps->i_mb_height >> h->sh.b_mbaff;
            int thread_mvy_range = i_fmv_range;

            if( h->param.i_threads > 1 )
            {
                int pix_y = (h->mb.i_mb_y | h->mb.b_interlaced) * 16;
                int thresh = pix_y + h->param.analyse.i_mv_range_thread;
                for( i = (h->sh.i_type == SLICE_TYPE_B); i >= 0; i-- )
                {
                    x264_frame_t **fref = i ? h->fref1 : h->fref0;
                    int i_ref = i ? h->i_ref1 : h->i_ref0;
                    for( j=0; j<i_ref; j++ )
                    {
                        x264_frame_cond_wait( fref[j], thresh );
                        thread_mvy_range = X264_MIN( thread_mvy_range, fref[j]->i_lines_completed - pix_y );
                    }
                }
                if( h->param.b_deterministic )
                    thread_mvy_range = h->param.analyse.i_mv_range_thread;
                if( h->mb.b_interlaced )
                    thread_mvy_range >>= 1;
            }

            h->mb.mv_min[1] = 4*( -16*mb_y - 24 );
            h->mb.mv_max[1] = 4*( 16*( mb_height - mb_y - 1 ) + 24 );
            h->mb.mv_min_spel[1] = x264_clip3( h->mb.mv_min[1], X264_MAX(4*(-512+i_spel_border), -i_fmv_range), i_fmv_range );
            h->mb.mv_max_spel[1] = CLIP_FMV( h->mb.mv_max[1] );
            h->mb.mv_max_spel[1] = X264_MIN( h->mb.mv_max_spel[1], thread_mvy_range*4 );
            h->mb.mv_min_fpel[1] = (h->mb.mv_min_spel[1]>>2) + i_fpel_border;
            h->mb.mv_max_fpel[1] = (h->mb.mv_max_spel[1]>>2) - i_fpel_border;
        }
#undef CLIP_FMV

        a->l0.me16x16.cost =
        a->l0.i_rd16x16    =
        a->l0.i_cost8x8    = COST_MAX;

        for( i = 0; i < 4; i++ )
        {
            a->l0.i_cost4x4[i] =
            a->l0.i_cost8x4[i] =
            a->l0.i_cost4x8[i] = COST_MAX;
        }

        a->l0.i_cost16x8   =
        a->l0.i_cost8x16   = COST_MAX;
        if( h->sh.i_type == SLICE_TYPE_B )
        {
            a->l1.me16x16.cost =
            a->l1.i_rd16x16    =
            a->l1.i_cost8x8    = COST_MAX;

            for( i = 0; i < 4; i++ )
            {
                a->l1.i_cost4x4[i] =
                a->l1.i_cost8x4[i] =
                a->l1.i_cost4x8[i] =
                a->i_cost8x8direct[i] = COST_MAX;
            }

            a->l1.i_cost16x8   =
            a->l1.i_cost8x16   =
            a->i_rd16x16bi     =
            a->i_rd16x16direct =
            a->i_rd8x8bi       =
            a->i_rd16x8bi      =
            a->i_rd8x16bi      =
            a->i_cost16x16bi   =
            a->i_cost16x16direct =
            a->i_cost8x8bi     =
            a->i_cost16x8bi    =
            a->i_cost8x16bi    = COST_MAX;
        }

        /* Fast intra decision */
        if( h->mb.i_mb_xy - h->sh.i_first_mb > 4 )
        {
            if(   IS_INTRA( h->mb.i_mb_type_left )
               || IS_INTRA( h->mb.i_mb_type_top )
               || IS_INTRA( h->mb.i_mb_type_topleft )
               || IS_INTRA( h->mb.i_mb_type_topright )
               || (h->sh.i_type == SLICE_TYPE_P && IS_INTRA( h->fref0[0]->mb_type[h->mb.i_mb_xy] ))
               || (h->mb.i_mb_xy - h->sh.i_first_mb < 3*(h->stat.frame.i_mb_count[I_4x4] + h->stat.frame.i_mb_count[I_8x8] + h->stat.frame.i_mb_count[I_16x16])) )
            { /* intra is likely */ }
            else
            {
                a->b_fast_intra = 1;
            }
        }
    }
}



/*
 * Handle intra mb
 */
/* Max = 4 */
static void predict_16x16_mode_available( unsigned int i_neighbour, int *mode, int *pi_count )
{
    if( i_neighbour & MB_TOPLEFT )
    {
        /* top and left available */
        *mode++ = I_PRED_16x16_V;
        *mode++ = I_PRED_16x16_H;
        *mode++ = I_PRED_16x16_DC;
        *mode++ = I_PRED_16x16_P;
        *pi_count = 4;
    }
    else if( i_neighbour & MB_LEFT )
    {
        /* left available*/
        *mode++ = I_PRED_16x16_DC_LEFT;
        *mode++ = I_PRED_16x16_H;
        *pi_count = 2;
    }
    else if( i_neighbour & MB_TOP )
    {
        /* top available*/
        *mode++ = I_PRED_16x16_DC_TOP;
        *mode++ = I_PRED_16x16_V;
        *pi_count = 2;
    }
    else
    {
        /* none available */
        *mode = I_PRED_16x16_DC_128;
        *pi_count = 1;
    }
}

/* Max = 4 */
static void predict_8x8chroma_mode_available( unsigned int i_neighbour, int *mode, int *pi_count )
{
    if( i_neighbour & MB_TOPLEFT )
    {