inter.c

T264是中国在JM系列和X264的基础上寻找的新的H264编程途径

/*****************************************************************************
 *
 *  T264 AVC CODEC
 *
 *  Copyright(C) 2004-2005 llcc <[e-mail]>
 *               2004-2005 visionany <[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  02111-1307 USA
 *
 ****************************************************************************/
#include "portab.h"
#include "stdio.h"
#ifndef CHIP_DM642
#include "memory.h"
#endif
#include "T264.h"
#include "inter.h"
#include "intra.h"
#include "estimation.h"
#include "utility.h"
#include "interpolate.h"
#include "bitstream.h"

//#define USE_PREV_DETECT

uint32_t
T264_predict_sad(T264_t* t, int32_t list)
{
    return T264_median(t->mb.sad_ref[0], t->mb.sad_ref[1], t->mb.sad_ref[2]);
}

void
T264_predict_mv(T264_t* t, int32_t list, int32_t i, int32_t width, T264_vector_t* vec)
{
    int32_t n;
    int32_t count = 0;
    int32_t idx;
    int32_t row;
    int32_t col;
    int32_t org;
    T264_vector_t vec_n[3];

    n = vec->refno;
    org = i;
    i = luma_index[i];

    col = org % 4;
    row = org / 4;

    vec_n[0] = t->mb.vec_ref[VEC_LUMA - 1 + row * 8 + col].vec[list];
    vec_n[1] = t->mb.vec_ref[VEC_LUMA - 8 + row * 8 + col].vec[list];
    vec_n[2] = t->mb.vec_ref[VEC_LUMA - 8 + row * 8 + col + width].vec[list];
    if (vec_n[2].refno == -2)
    {
        vec_n[2] = t->mb.vec_ref[VEC_LUMA - 8 + row * 8 + col - 1].vec[list];
    }
    if (((i & 3) == 3) || ((i & 3) == 2 && width == 2))
    {
        vec_n[2] = t->mb.vec_ref[VEC_LUMA - 8 + row * 8 + col - 1].vec[list];
    }
    if (t->mb.mb_part == MB_16x8)
    {
        if (i == 0 && n == vec_n[1].refno)
        {
            vec[0].x = vec_n[1].x;
            vec[0].y = vec_n[1].y;
            return;
        }
        else if (i != 0 && n == vec_n[0].refno)
        {
            vec[0].x = vec_n[0].x;
            vec[0].y = vec_n[0].y;
            return;
        }
    }
    else if (t->mb.mb_part == MB_8x16)
    {
        if (i == 0 && n == vec_n[0].refno)
        {
            vec[0].x = vec_n[0].x;
            vec[0].y = vec_n[0].y;
            return;
        }
        else if (i != 0 && n == vec_n[2].refno)
        {
            vec[0].x = vec_n[2].x;
            vec[0].y = vec_n[2].y;
            return;
        }
    }

    if (vec_n[0].refno == n)
    {
        count ++;
        idx = 0;
    }
    if (vec_n[1].refno == n)
    {
        count ++;
        idx = 1;
    }
    if (vec_n[2].refno == n)
    {
        count ++;
        idx = 2;
    }

    if (count > 1)
    {
        vec[0].x = T264_median(vec_n[0].x, vec_n[1].x, vec_n[2].x);
        vec[0].y = T264_median(vec_n[0].y, vec_n[1].y, vec_n[2].y);
        return;
    }
    else if (count == 1)
    {
        vec[0].x = vec_n[idx].x;
        vec[0].y = vec_n[idx].y;
        return;
    }
    else if (vec_n[1].refno == -2 && vec_n[2].refno == -2 && vec_n[0].refno != -2)
    {
        vec[0].x = vec_n[0].x;
        vec[0].y = vec_n[0].y;
    }
    else
    {
        vec[0].x = T264_median(vec_n[0].x, vec_n[1].x, vec_n[2].x);
        vec[0].y = T264_median(vec_n[0].y, vec_n[1].y, vec_n[2].y);
        return;        
    }
}

void
T264_predict_mv_skip(T264_t* t, int32_t list, T264_vector_t* vec)
{
    T264_vector_t vec_n[2];
    int32_t zero_left, zero_top;

    vec_n[0] = t->mb.vec_ref[VEC_LUMA - 1].vec[0];
    vec_n[1] = t->mb.vec_ref[VEC_LUMA - 8].vec[0];
    vec[0].refno = 0;

    zero_left = vec_n[0].refno == -2 ? 1 : vec_n[0].refno == 0 && vec_n[0].x == 0 && vec_n[0].y == 0 ? 1 : 0;
    zero_top  = vec_n[1].refno == -2 ? 1 : vec_n[1].refno == 0 && vec_n[1].x == 0 && vec_n[1].y == 0 ? 1 : 0;

    if (zero_left || zero_top)
    {
        vec[0].x = vec[0].y = 0;
    }
    else
    {
        T264_predict_mv(t, 0, 0, 4, vec);
    }
}

int32_t 
T264_median(int32_t x, int32_t y, int32_t z)
{
    int32_t min, max;
    if (x < y)
    {
        min = x;
        max = y;
    }
    else
    {
        min = y;
        max = x;
    }

    if (z < min)
    {
        min = z;
    }
    else if (z > max)
    {
        max = z;
    }

    return x + y + z - min - max;
}

void
copy_nvec(T264_vector_t* src, T264_vector_t* dst, int32_t width, int32_t height, int32_t stride)
{
    int32_t i, j;

    for(i = 0 ; i < height ; i ++)
    {
        for(j = 0 ; j < width ; j ++)
        {
            dst[j] = src[0];
        }
        dst += stride;
    }
}

void
T264_inter_p16x16_mode_available(T264_t* t, int32_t preds[], int32_t* modes)
{
    if (t->flags & USE_FORCEBLOCKSIZE)
    {
        *modes = 0;
      
        preds[(*modes) ++] = MB_16x16;

        if (t->param.block_size & SEARCH_16x8P)
            preds[(*modes) ++] = MB_16x8;
        if (t->param.block_size & SEARCH_8x16P)
            preds[(*modes) ++] = MB_8x16;

        return ;
    }

    if ((t->mb.mb_neighbour & (MB_LEFT | MB_TOP)) == (MB_LEFT | MB_TOP))
    {
        *modes = 0;

        preds[(*modes) ++] = MB_16x16;
        if (t->mb.vec_ref[VEC_LUMA - 1].part == MB_16x8)
        {
            preds[(*modes) ++] = MB_16x8;
        }
        if (t->mb.vec_ref[VEC_LUMA - 8].part == MB_8x16)
        {
            preds[(*modes) ++] = MB_8x16;
        }
    }
    else
    {
        // try all
        preds[0] = MB_16x16;
        preds[1] = MB_16x8;
        preds[2] = MB_8x16;
        *modes = 3;
    }
}

void
T264_inter_p8x8_mode_available(T264_t* t, int32_t preds[], int32_t* modes, int32_t sub_no)
{
    static const int32_t neighbour[] = 
    {
        0, MB_LEFT, MB_TOP, MB_LEFT| MB_TOP
    };

    int32_t mb_neighbour = t->mb.mb_neighbour| neighbour[sub_no];

    if (t->flags & USE_FORCEBLOCKSIZE)
    {
        *modes = 0;

        if (t->param.block_size & SEARCH_8x8P)
            preds[(*modes) ++] = MB_8x8;
        if (t->param.block_size & SEARCH_8x4P)
            preds[(*modes) ++] = MB_8x4;
        if (t->param.block_size & SEARCH_4x8P)
            preds[(*modes) ++] = MB_4x8;
        if (t->param.block_size & SEARCH_4x4P)
            preds[(*modes) ++] = MB_4x4;

        return ;
    }

    if ((mb_neighbour & (MB_LEFT | MB_TOP)) == (MB_LEFT | MB_TOP))
    {
        *modes = 0;

        preds[*modes ++] = MB_8x8;
        if (t->mb.vec_ref[VEC_LUMA - 8 + sub_no / 2 * 16 + sub_no % 2 * 4].part == MB_8x4)
        {
            preds[*modes ++] = MB_8x4;
        }
        if (t->mb.vec_ref[VEC_LUMA - 1 + sub_no / 2 * 16 + sub_no % 2 * 4].part == MB_4x8)
        {
            preds[*modes ++] = MB_4x8;
        }
        if (t->mb.vec_ref[VEC_LUMA - 8 + sub_no / 2 * 16 + sub_no % 2 * 4].part == MB_4x4 || 
            t->mb.vec_ref[VEC_LUMA - 1 + sub_no / 2 * 16 + sub_no % 2 * 4].part == MB_4x4)
        {
            preds[*modes ++] = MB_4x4;
        }
    }
    else
    {
        // try all
        preds[0] = MB_8x8;
        preds[1] = MB_8x4;
        preds[2] = MB_4x8;
        preds[3] = MB_4x4;
        *modes = 4;
    }
}

int32_t
T264_get_pos_sad(T264_t* t, uint8_t* ref, T264_vector_t* vec)
{
    uint8_t* tmp;
    int32_t x, y;
    int32_t list_index = 0;
    uint32_t sad;

    static const int8_t index[4][4][6] = 
    {
        {{0, 0, 0, 0, 0, 0}, {0, 1, 0, 0, 0, 0}, {1, 1, 0, 0, 0, 0}, {1, 0, 0, 0, 1, 0}},
        {{0, 2, 0, 0, 0, 0}, {1, 2, 0, 0, 0, 0}, {1, 3, 0, 0, 0, 0}, {1, 2, 0, 0, 1, 0}},
        {{2, 2, 0, 0, 0, 0}, {2, 3, 0, 0, 0, 0}, {3, 3, 0, 0, 0, 0}, {3, 2, 0, 0, 1, 0}},
        {{2, 0, 0, 0, 0, 1}, {2, 1, 0, 0, 0, 1}, {3, 1, 0, 0, 0, 1}, {1, 2, 0, 1, 1, 0}}
    };

    // need subpel
    if ((t->flags & (USE_QUARTPEL)) == (USE_QUARTPEL))
    {
        x = (vec->x & 3);
        y = (vec->y & 3);

        if (index[y][x][0] == index[y][x][1])
        {
            tmp = t->ref[list_index][vec->refno]->Y[index[y][x][0]] + ((t->mb.mb_y << 4) + (vec->y >> 2)) * t->edged_stride + 
                ((t->mb.mb_x << 4) + (vec->x >> 2));
            t->memcpy_stride_u(tmp, 16, 16, t->edged_stride, ref, 16);
        }
        else
        {
            t->pia[MB_16x16](t->ref[list_index][vec->refno]->Y[index[y][x][0]] + ((t->mb.mb_y << 4) + (vec->y >> 2) + index[y][x][3]) * t->edged_stride + (t->mb.mb_x << 4) + (vec->x >> 2) + index[y][x][2], 
                t->ref[list_index][vec->refno]->Y[index[y][x][1]] + ((t->mb.mb_y << 4) + (vec->y >> 2) + index[y][x][5]) * t->edged_stride + (t->mb.mb_x << 4) + (vec->x >> 2) + index[y][x][4],
                t->edged_stride, t->edged_stride, ref, 16);
        }

        // rc will use sad value
        sad = t->cmp[MB_16x16](t->mb.src_y, t->stride, ref, 16);

        return sad;
    }

    return -1;
}

int32_t
T264_detect_pskip(T264_t* t, uint32_t sad_t)
{
    // detect p skip has a two-step process. here try to find suitable skip mv
    //  and encode post will decide if use skip mode or not
    DECLARE_ALIGNED_MATRIX(ref, 16, 16, uint8_t, CACHE_SIZE);

    T264_vector_t vec;
    uint8_t* tmp;
    int32_t x, y;
    int32_t i, j;
    int32_t list_index = 0;

    static const int8_t index[4][4][6] = 
    {
        {{0, 0, 0, 0, 0, 0}, {0, 1, 0, 0, 0, 0}, {1, 1, 0, 0, 0, 0}, {1, 0, 0, 0, 1, 0}},
        {{0, 2, 0, 0, 0, 0}, {1, 2, 0, 0, 0, 0}, {1, 3, 0, 0, 0, 0}, {1, 2, 0, 0, 1, 0}},
        {{2, 2, 0, 0, 0, 0}, {2, 3, 0, 0, 0, 0}, {3, 3, 0, 0, 0, 0}, {3, 2, 0, 0, 1, 0}},
        {{2, 0, 0, 0, 0, 1}, {2, 1, 0, 0, 0, 1}, {3, 1, 0, 0, 0, 1}, {1, 2, 0, 1, 1, 0}}
    };

    T264_predict_mv_skip(t, 0, &vec);

    // need subpel
    if ((t->flags & (USE_QUARTPEL)) == (USE_QUARTPEL))
    {
        x = (vec.x & 3);
        y = (vec.y & 3);

        if (index[y][x][0] == index[y][x][1])
        {
            tmp = t->ref[list_index][vec.refno]->Y[index[y][x][0]] + ((t->mb.mb_y << 4) + (vec.y >> 2)) * t->edged_stride + 
                ((t->mb.mb_x << 4) + (vec.x >> 2));
            t->memcpy_stride_u(tmp, 16, 16, t->edged_stride, ref, 16);
        }
        else
        {
            t->pia[MB_16x16](t->ref[list_index][vec.refno]->Y[index[y][x][0]] + ((t->mb.mb_y << 4) + (vec.y >> 2) + index[y][x][3]) * t->edged_stride + (t->mb.mb_x << 4) + (vec.x >> 2) + index[y][x][2], 
                        t->ref[list_index][vec.refno]->Y[index[y][x][1]] + ((t->mb.mb_y << 4) + (vec.y >> 2) + index[y][x][5]) * t->edged_stride + (t->mb.mb_x << 4) + (vec.x >> 2) + index[y][x][4],
                        t->edged_stride, t->edged_stride, ref, 16);