x264 - predict.c

最新推荐文章于 2021-10-27 20:54:27 发布
置顶 最新推荐文章于 2021-10-27 20:54:27 发布
阅读量1.3k 收藏 1
点赞数
分类专栏： Stream Media/H264
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/fanbird2008/article/details/9158843
版权


/****************************************************************************

 * 16x16 prediction for intra luma block
 ****************************************************************************/


#define PREDICT_16x16_DC(v)\
    for( int i = 0; i < 16; i++ )\
    {\
        MPIXEL_X4( src+ 0 ) = v;\
        MPIXEL_X4( src+ 4 ) = v;\
        MPIXEL_X4( src+ 8 ) = v;\
        MPIXEL_X4( src+12 ) = v;\
        src += FDEC_STRIDE;\
    }


void x264_predict_16x16_dc_c( pixel *src )
{
    int dc = 0;

/*

    dc = H0,-1 + ... + H15,-1 + V-1,0 + ... + V-1,15

    dc = (dc + 16） / 32

    用dc组合成一个32bits字（即dcsplat的第一个字节是dc, 第二个字节也是dc, 第三个字节也是dc, 第四个字节也是dc） 

    PREDICT_16x16_DC给16行赋值，每行按4个32bits字赋值，都赋值为dcsplat;

*/

    for( int i = 0; i < 16; i++ )
    {
        dc += src[-1 + i * FDEC_STRIDE];
        dc += src[i - FDEC_STRIDE];
    }
    pixel4 dcsplat = PIXEL_SPLAT_X4( ( dc + 16 ) >> 5 );


    PREDICT_16x16_DC( dcsplat );
}


static void x264_predict_16x16_dc_left_c( pixel *src )
{
    int dc = 0;

/*

    dc = V-1,0 + ... + V-1,15

    dc = (dc + 8) / 16

    用dc组合成一个32bits字（即dcsplat的第一个字节是dc, 第二个字节也是dc, 第三个字节也是dc, 第四个字节也是dc） 

    PREDICT_16x16_DC给16行赋值，每行按4个32bits字赋值，都赋值为dcsplat;

*/
    for( int i = 0; i < 16; i++ )
        dc += src[-1 + i * FDEC_STRIDE];
    pixel4 dcsplat = PIXEL_SPLAT_X4( ( dc + 8 ) >> 4 );


    PREDICT_16x16_DC( dcsplat );
}


static void x264_predict_16x16_dc_top_c( pixel *src )
{
    int dc = 0;

/*

    dc = H0,-1 + ... + H15,-1

    dc = (dc + 8) / 16

    用dc组合成一个32bits字（即dcsplat的第一个字节是dc, 第二个字节也是dc, 第三个字节也是dc, 第四个字节也是dc） 

    PREDICT_16x16_DC给16行赋值，每行按4个32bits字赋值，都赋值为dcsplat;

*/

    for( int i = 0; i < 16; i++ )
        dc += src[i - FDEC_STRIDE];
    pixel4 dcsplat = PIXEL_SPLAT_X4( ( dc + 8 ) >> 4 );


    PREDICT_16x16_DC( dcsplat );
}


static void x264_predict_16x16_dc_128_c( pixel *src )
{

/*

  将16行16列的每个字节赋值为128

*/


    PREDICT_16x16_DC( PIXEL_SPLAT_X4( 1 << (BIT_DEPTH-1) ) );
}


void x264_predict_16x16_h_c( pixel *src )
{

/*

将src[-1]字节，延展为32bits字v, v的每个字节都是src[-1]

然后将每一行的4个32bits字都赋值为v

*/
    for( int i = 0; i < 16; i++ )
    {
        const pixel4 v = PIXEL_SPLAT_X4( src[-1] );
        MPIXEL_X4( src+ 0 ) = v;
        MPIXEL_X4( src+ 4 ) = v;
        MPIXEL_X4( src+ 8 ) = v;
        MPIXEL_X4( src+12 ) = v;
        src += FDEC_STRIDE;
    }
}
void x264_predict_16x16_v_c( pixel *src )
{

/*

将  -1 行的4个32bits的字分别赋值给v0, v1, v2, v3

将16行中的每一行的4个32bits字， 分别赋值为v0, v1, v2, v3

*/
    pixel4 v0 = MPIXEL_X4( &src[ 0-FDEC_STRIDE] );
    pixel4 v1 = MPIXEL_X4( &src[ 4-FDEC_STRIDE] );
    pixel4 v2 = MPIXEL_X4( &src[ 8-FDEC_STRIDE] );
    pixel4 v3 = MPIXEL_X4( &src[12-FDEC_STRIDE] );


    for( int i = 0; i < 16; i++ )
    {
        MPIXEL_X4( src+ 0 ) = v0;
        MPIXEL_X4( src+ 4 ) = v1;
        MPIXEL_X4( src+ 8 ) = v2;
        MPIXEL_X4( src+12 ) = v3;
        src += FDEC_STRIDE;
    }
}


void x264_predict_16x16_p_c( pixel *src )
{
    int H = 0, V = 0;


    /* calculate H and V */

/*

H =  H8,-1 - H6,-1 + 2(H9,-1 - H5,-1) + 3(H10,-1 - H4,-1) + 4(H11,-1 - H3,-1)

    + 5(H12,-1 - H2,-1) + 6(H13,-1 - H1,-1) + 7(H14,-1 - H0,-1) + 8(H15,-1 - H-1,-1)

V = V-1,8 - V-1,6 + 2(V-1,9 - V-1,5) +3(V-1,10 - V-1,4) + 4(V-1,11 - V-1,3)

   + 5(V-1,12 - V-1,2) + 6(V-1,13 - V-1,1) + 7(V-1,14 - V-1,0) + 8(V-1,15 - V-1,-1)




    H + V =    H8,-1 + V-1,8 - (H6,-1 + V-1,6)

              + 2(H9,-1 + V-1,9) - 2(H5,-1 + V-1,5) 

              + 3(H10,-1 + V-1,10) - 3(H4,-1 + V-1,4)

              + 4(H11,-1 + V-1,11) - 4(H3,-1 + V-1,3)

              + 5(H12,-1 + V-1,12) - 5(H2,-1 + V-1,2)

              + 6(H13,-1 + V-1,13) - 6(H1,-1 + V-1,1)

              + 7(H14,-1 + V-1,14) - 7(H0,-1 + V-1,0)

              + 8(H15,-1 + V-1,15) - 8(H-1,-1 + V-1,-1) 

  */

     

    for( int i = 0; i <= 7; i++ )
    {
        H += ( i + 1 ) * ( src[ 8 + i - FDEC_STRIDE ] - src[6 -i -FDEC_STRIDE] );
        V += ( i + 1 ) * ( src[-1 + (8+i)*FDEC_STRIDE] - src[-1 + (6-i)*FDEC_STRIDE] );
    }


/*

a = 16(H15,-1 + V-1,15)

b = (5H + 32) / 64

c = (5V + 32) / 64

*/   

    int a = 16 * ( src[-1 + 15*FDEC_STRIDE] + src[15 - FDEC_STRIDE] );
    int b = ( 5 * H + 32 ) >> 6;
    int c = ( 5 * V + 32 ) >> 6;


    int i00 = a - b * 7 - c * 7 + 16;


    for( int y = 0; y < 16; y++ )
    {
        int pix = i00;
        for( int x = 0; x < 16; x++ )
        {
            src[x] = x264_clip_pixel( pix>>5 );
            pix += b;
        }
        src += FDEC_STRIDE;
        i00 += c;
    }
}




/****************************************************************************
 * 8x8 prediction for intra chroma block (4:2:0)
 ****************************************************************************/


static void x264_predict_8x8c_dc_128_c( pixel *src )
{

/*

将每一行的每个字节都赋值为128

*/
    for( int y = 0; y < 8; y++ )
    {
        MPIXEL_X4( src+0 ) = PIXEL_SPLAT_X4( 1 << (BIT_DEPTH-1) );
        MPIXEL_X4( src+4 ) = PIXEL_SPLAT_X4( 1 << (BIT_DEPTH-1) );
        src += FDEC_STRIDE;
    }
}
static void x264_predict_8x8c_dc_left_c( pixel *src )
{
    int dc0 = 0, dc1 = 0;
/*

dc0 = V-1,0 + ... + V-1,3 

dc1 = V-1,4 + ... + V-1,7

dc0 = (dc0 + 2)  / 4;

dc1 = (dc1 + 2) / 4;

dc0splat = dc0延展为32bits字， 每个字节=dc0

dc0splat = dc1延展为32bits字， 每个字节=dc1

将 0 - 3 行的每一行的2个32bits字都赋值为dc0splat

将 4 - 7 行的每一行的2个32bits字都赋值为dc1splat

*/

    for( int y = 0; y < 4; y++ )
    {
        dc0 += src[y * FDEC_STRIDE     - 1];
        dc1 += src[(y+4) * FDEC_STRIDE - 1];
    }
    pixel4 dc0splat = PIXEL_SPLAT_X4( ( dc0 + 2 ) >> 2 );
    pixel4 dc1splat = PIXEL_SPLAT_X4( ( dc1 + 2 ) >> 2 );


    for( int y = 0; y < 4; y++ )
    {
        MPIXEL_X4( src+0 ) = dc0splat;
        MPIXEL_X4( src+4 ) = dc0splat;
        src += FDEC_STRIDE;
    }
    for( int y = 0; y < 4; y++ )
    {
        MPIXEL_X4( src+0 ) = dc1splat;
        MPIXEL_X4( src+4 ) = dc1splat;
        src += FDEC_STRIDE;
    }


}
static void x264_predict_8x8c_dc_top_c( pixel *src )
{
    int dc0 = 0, dc1 = 0;

/*

dc0 = H0,-1 + ... + H3,-1 

dc1 = H4,-1 + ... + H7,-1

dc0 = (dc0 + 2)  / 4;

dc1 = (dc1 + 2) / 4;

dc0splat = dc0延展为32bits字， 每个字节=dc0

dc0splat = dc1延展为32bits字， 每个字节=dc1

将 0 - 7 行的每一行的2个32bits字的第一个字赋值为dc0splat， 第二个字赋值为dc1splat

*/

    for( int x = 0; x < 4; x++ )
    {
        dc0 += src[x     - FDEC_STRIDE];
        dc1 += src[x + 4 - FDEC_STRIDE];
    }
    pixel4 dc0splat = PIXEL_SPLAT_X4( ( dc0 + 2 ) >> 2 );
    pixel4 dc1splat = PIXEL_SPLAT_X4( ( dc1 + 2 ) >> 2 );


    for( int y = 0; y < 8; y++ )
    {
        MPIXEL_X4( src+0 ) = dc0splat;
        MPIXEL_X4( src+4 ) = dc1splat;
        src += FDEC_STRIDE;
    }
}


void x264_predict_8x8c_dc_c( pixel *src )
{
    int s0 = 0, s1 = 0, s2 = 0, s3 = 0;


    /*
       s0 = H0,-1 + ... + H3,-1

       s1 = H4,-1 + ... + H7,-1
       s2 = V-1,0 + ... + V-1,3
       s3 = V-1,4 + ... + V-1,7
    */


    for( int i = 0; i < 4; i++ )
    {
        s0 += src[i - FDEC_STRIDE];
        s1 += src[i + 4 - FDEC_STRIDE];
        s2 += src[-1 + i * FDEC_STRIDE];
        s3 += src[-1 + (i+4)*FDEC_STRIDE];
    }
    /*
       dc0 = (s0 + s2 + 4) / 8， dc0 延展

       dc1 = (s1 + 2) / 4，  dc1 延展
       dc2 = (s3 + 2) / 4，   dc2 延展

       dc3 = (s1 + s3 + 4) / 8， dc3  延展
     */
    pixel4 dc0 = PIXEL_SPLAT_X4( ( s0 + s2 + 4 ) >> 3 );
    pixel4 dc1 = PIXEL_SPLAT_X4( ( s1 + 2 ) >> 2 );
    pixel4 dc2 = PIXEL_SPLAT_X4( ( s3 + 2 ) >> 2 );
    pixel4 dc3 = PIXEL_SPLAT_X4( ( s1 + s3 + 4 ) >> 3 );

/*

将 0 - 3行的2个32bits字分别赋值为dc0, dc1

*/
    for( int y = 0; y < 4; y++ )
    {
        MPIXEL_X4( src+0 ) = dc0;
        MPIXEL_X4( src+4 ) = dc1;
        src += FDEC_STRIDE;
    }


/*

将 4 - 7行的2个32bits字分别赋值为dc2, dc3

*/

    for( int y = 0; y < 4; y++ )
    {
        MPIXEL_X4( src+0 ) = dc2;
        MPIXEL_X4( src+4 ) = dc3;
        src += FDEC_STRIDE;
    }
}
void x264_predict_8x8c_h_c( pixel *src )
{

/*

将每一行的src[-1] 赋值给这行的每一个字节

*/


    for( int i = 0; i < 8; i++ )
    {
        pixel4 v = PIXEL_SPLAT_X4( src[-1] );
        MPIXEL_X4( src+0 ) = v;
        MPIXEL_X4( src+4 ) = v;
        src += FDEC_STRIDE;
    }
}
void x264_predict_8x8c_v_c( pixel *src )
{

/*

将  -1 行 的 2 个 32bits的字 分别 赋值给 v0, v1

将8行中的每一行的2个32bits字分别赋值为v0, v1

*/
    pixel4 v0 = MPIXEL_X4( src+0-FDEC_STRIDE );
    pixel4 v1 = MPIXEL_X4( src+4-FDEC_STRIDE );


    for( int i = 0; i < 8; i++ )
    {
        MPIXEL_X4( src+0 ) = v0;
        MPIXEL_X4( src+4 ) = v1;
        src += FDEC_STRIDE;
    }
}
void x264_predict_8x8c_p_c( pixel *src )
{
    int H = 0, V = 0;

/*

H = H4,-1 - H2,-1 + 2(H5,-1 - H1,-1) + 3(H6,-1 - H0,-1) + 4(H7,-1 - H-1,-1)

V = V-1,4 - V-1,2 + 2(V-1,5 - V-1,1) + 3(V-1,6 - V-1,0) + 4(V-1,7 - V-1,-1)

*/


    for( int i = 0; i < 4; i++ )
    {
        H += ( i + 1 ) * ( src[4+i - FDEC_STRIDE] - src[2 - i -FDEC_STRIDE] );
        V += ( i + 1 ) * ( src[-1 +(i+4)*FDEC_STRIDE] - src[-1+(2-i)*FDEC_STRIDE] );
    }

/*

a = 16(V-1,7 + H7,-1)

b = (17H + 16) / 32

c = (17V + 16) / 32

i00 = a - 3b - 3c + 16

*/

    int a = 16 * ( src[-1+7*FDEC_STRIDE] + src[7 - FDEC_STRIDE] );
    int b = ( 17 * H + 16 ) >> 5;
    int c = ( 17 * V + 16 ) >> 5;
    int i00 = a -3*b -3*c + 16;


    for( int y = 0; y < 8; y++ )
    {
        int pix = i00;
        for( int x = 0; x < 8; x++ )
        {
            src[x] = x264_clip_pixel( pix>>5 );
            pix += b;
        }
        src += FDEC_STRIDE;
        i00 += c;
    }
}


/****************************************************************************
 * 8x16 prediction for intra chroma block (4:2:2)
 ****************************************************************************/


static void x264_predict_8x16c_dc_128_c( pixel *src )
{

/*

将16行的每一行的每一个字节都赋值为128

*/


    for( int y = 0; y < 16; y++ )
    {
        MPIXEL_X4( src+0 ) = PIXEL_SPLAT_X4( 1 << (BIT_DEPTH-1) );
        MPIXEL_X4( src+4 ) = PIXEL_SPLAT_X4( 1 << (BIT_DEPTH-1) );
        src += FDEC_STRIDE;
    }
}


static void x264_predict_8x16c_dc_left_c( pixel *src )
{

/*

将16行以4行为一个单位， 对这个四行求dc

dc = V-1,0 + ... + V-1,3

dc = (dc + 2) / 4

dc延展为dcsplat

将dcsplat赋值给这四行中的每一行

*/


    for( int i = 0; i < 4; i++ )
    {
        int dc = 0;


        for( int y = 0; y < 4; y++ )
            dc += src[y*FDEC_STRIDE - 1];


        pixel4 dcsplat = PIXEL_SPLAT_X4( (dc + 2) >> 2 );


        for( int y = 0; y < 4; y++ )
        {
            MPIXEL_X4( src+0 ) = dcsplat;
            MPIXEL_X4( src+4 ) = dcsplat;
            src += FDEC_STRIDE;
        }
    }
}


static void x264_predict_8x16c_dc_top_c( pixel *src )
{
    int dc0 = 0, dc1 = 0;

/*

dc0= H0,-1 + ... + H3,-1

dc1= H4,-1 + ... + H7,-1

dc0 = (dc0 + 2) / 4

dc1 = (dc1 + 2) / 4

dc0延展为dc0splat, dc1 延展为dc1splat

将dcsplat0, dcsplat1赋值给16行中的每一行

*/

    for(int  x = 0; x < 4; x++ )
    {
        dc0 += src[x     - FDEC_STRIDE];
        dc1 += src[x + 4 - FDEC_STRIDE];
    }
    pixel4 dc0splat = PIXEL_SPLAT_X4( ( dc0 + 2 ) >> 2 );
    pixel4 dc1splat = PIXEL_SPLAT_X4( ( dc1 + 2 ) >> 2 );


    for( int y = 0; y < 16; y++ )
    {
        MPIXEL_X4( src+0 ) = dc0splat;
        MPIXEL_X4( src+4 ) = dc1splat;
        src += FDEC_STRIDE;
    }
}


void x264_predict_8x16c_dc_c( pixel *src )
{
    int s0 = 0, s1 = 0, s2 = 0, s3 = 0, s4 = 0, s5 = 0;


    /*
          s0 s1
       s2
       s3
       s4
       s5
    */

/*

s0 = H0,-1 + ... + H3,-1

s1 = H4,-1 + ... + H7,-1

s2 = V-1,0 + ... + V-1,3

s3 = V-1,4 + ... + V-1,7

s4 = V-1,8 + ... + V-1,11

s5 = V-1,12 + ... + V-1,15,

*/


    for( int i = 0; i < 4; i++ )
    {
        s0 += src[i+0 - FDEC_STRIDE];
        s1 += src[i+4 - FDEC_STRIDE];
        s2 += src[-1 + (i+0)  * FDEC_STRIDE];
        s3 += src[-1 + (i+4)  * FDEC_STRIDE];
        s4 += src[-1 + (i+8)  * FDEC_STRIDE];
        s5 += src[-1 + (i+12) * FDEC_STRIDE];
    }
    /*
       dc0 dc1
       dc2 dc3
       dc4 dc5
       dc6 dc7
    */

/*

dc0 = (s0 + s2 + 4) / 8

dc1 = (s1 + 2) / 4

dc2 = (s3 + 2) / 4

dc3 = (s1 + s3 + 4) / 8

dc4 = (s4 + 2) / 4

dc5 = (s1 + s4 + 4) / 8

dc6 = (s5 + 2) / 4

dc7 = (s1 + s5 + 4) / 8

将dc0延展， dc1延展， 然后赋值给0-3行

dc2, dc3分别延展，       然后赋值给4-7行

dc4, dc5分别延展，       然后赋值给8-11行

dc6, dc7分别延展，      然后赋值给12-15行

*/


    pixel4 dc0 = PIXEL_SPLAT_X4( ( s0 + s2 + 4 ) >> 3 );
    pixel4 dc1 = PIXEL_SPLAT_X4( ( s1 + 2 ) >> 2 );
    pixel4 dc2 = PIXEL_SPLAT_X4( ( s3 + 2 ) >> 2 );
    pixel4 dc3 = PIXEL_SPLAT_X4( ( s1 + s3 + 4 ) >> 3 );
    pixel4 dc4 = PIXEL_SPLAT_X4( ( s4 + 2 ) >> 2 );
    pixel4 dc5 = PIXEL_SPLAT_X4( ( s1 + s4 + 4 ) >> 3 );
    pixel4 dc6 = PIXEL_SPLAT_X4( ( s5 + 2 ) >> 2 );
    pixel4 dc7 = PIXEL_SPLAT_X4( ( s1 + s5 + 4 ) >> 3 );


    for( int y = 0; y < 4; y++ )
    {
        MPIXEL_X4( src+0 ) = dc0;
        MPIXEL_X4( src+4 ) = dc1;
        src += FDEC_STRIDE;
    }
    for( int y = 0; y < 4; y++ )
    {
        MPIXEL_X4( src+0 ) = dc2;
        MPIXEL_X4( src+4 ) = dc3;
        src += FDEC_STRIDE;
    }
    for( int y = 0; y < 4; y++ )
    {
        MPIXEL_X4( src+0 ) = dc4;
        MPIXEL_X4( src+4 ) = dc5;
        src += FDEC_STRIDE;
    }
    for( int y = 0; y < 4; y++ )
    {
        MPIXEL_X4( src+0 ) = dc6;
        MPIXEL_X4( src+4 ) = dc7;
        src += FDEC_STRIDE;
    }
}
void x264_predict_8x16c_h_c( pixel *src )
{

/*

将每一行的src[-1]  赋值给这行的每一个字节

*/


    for( int i = 0; i < 16; i++ )
    {
        pixel4 v = PIXEL_SPLAT_X4( src[-1] );
        MPIXEL_X4( src+0 ) = v;
        MPIXEL_X4( src+4 ) = v;
        src += FDEC_STRIDE;
    }
}


void x264_predict_8x16c_v_c( pixel *src )
{
    pixel4 v0 = MPIXEL_X4( src+0-FDEC_STRIDE );
    pixel4 v1 = MPIXEL_X4( src+4-FDEC_STRIDE );

/*

将 -1 行的 2 个 32 bits 的 字 分别赋值给v0， v1

然后将v0, v1赋值给16行中的每一行

*/

    for( int i = 0; i < 16; i++ )
    {
        MPIXEL_X4( src+0 ) = v0;
        MPIXEL_X4( src+4 ) = v1;
        src += FDEC_STRIDE;
    }
}


void x264_predict_8x16c_p_c( pixel *src )
{
    int H = 0;
    int V = 0;

/*

H = H4,-1 - H2,-1 + 2(H5,-1 - H1,-1) + 3(H6,-1 - H0,-1) + 4(H7,-1 - H-1,-1)

V = V-1,8 - V-1,6 + 2(V-1,9 - V-1,5) + 3(V-1,10 - V-1,4) + 4(V-1,11 - V-1,3) + 5(V-1,12 - V-1,2) + 6(V-1,13 - V-1,1) + 7(V-1,14 - V-1,0) + 8(V-1,15 - V-1,-1) 

*/


    for( int i = 0; i < 4; i++ )
        H += ( i + 1 ) * ( src[4 + i - FDEC_STRIDE] - src[2 - i - FDEC_STRIDE] );
    for( int i = 0; i < 8; i++ )
        V += ( i + 1 ) * ( src[-1 + (i+8)*FDEC_STRIDE] - src[-1 + (6-i)*FDEC_STRIDE] );

/*

a = 16(H7,-1 + V-1,15)

b = (17H + 16) / 32

c = (5V + 32) / 64

i00 = a - 3b -7c + 16

*/

    int a = 16 * ( src[-1 + 15*FDEC_STRIDE] + src[7 - FDEC_STRIDE] );
    int b = ( 17 * H + 16 ) >> 5;
    int c = ( 5 * V + 32 ) >> 6;
    int i00 = a -3*b -7*c + 16;


    for( int y = 0; y < 16; y++ )
    {
        int pix = i00;
        for( int x = 0; x < 8; x++ )
        {
            src[x] = x264_clip_pixel( pix>>5 );
            pix += b;
        }
        src += FDEC_STRIDE;
        i00 += c;
    }
}


/****************************************************************************
 * 4x4 prediction for intra luma block
 ****************************************************************************/


#define SRC(x,y) src[(x)+(y)*FDEC_STRIDE]        // src[y][x]

#define SRC_X4(x,y) MPIXEL_X4( &SRC(x,y) )      // (uint32_t *)&src[y][x]


//  SRC_X4(0, 0)  denote 0 row of 4x4

//  SRC_X4(0, 1) denote 1 row of 4x4

//  SRC_X4(0, 2) denote 2 row of 4x4

// SRC_X4(0, 3) denote 3 row of 4x4

#define PREDICT_4x4_DC(v)\
    SRC_X4(0,0) = SRC_X4(0,1) = SRC_X4(0,2) = SRC_X4(0,3) = v;

// mode 2 (DC), but H and V both no available
static void x264_predict_4x4_dc_128_c( pixel *src )
{
    PREDICT_4x4_DC( PIXEL_SPLAT_X4( 1 << (BIT_DEPTH-1) ) );     // all = 128

}

// mode 2(DC), but H no available

static void x264_predict_4x4_dc_left_c( pixel *src )
{
    pixel4 dc = PIXEL_SPLAT_X4( (SRC(-1,0) + SRC(-1,1) + SRC(-1,2) + SRC(-1,3) + 2) >> 2 );
    PREDICT_4x4_DC( dc );

}

// mode 2(DC), but V no available

static void x264_predict_4x4_dc_top_c( pixel *src )
{
    pixel4 dc = PIXEL_SPLAT_X4( (SRC(0,-1) + SRC(1,-1) + SRC(2,-1) + SRC(3,-1) + 2) >> 2 );
    PREDICT_4x4_DC( dc );

}

// mode 2(DC), H and V both available

void x264_predict_4x4_dc_c( pixel *src )
{
    pixel4 dc = PIXEL_SPLAT_X4( (SRC(-1,0) + SRC(-1,1) + SRC(-1,2) + SRC(-1,3) +
                                 SRC(0,-1) + SRC(1,-1) + SRC(2,-1) + SRC(3,-1) + 4) >> 3 );
    PREDICT_4x4_DC( dc );

}


// mode 1 (Horizon), take left byte and splat_x4 to 4 byte, then copy current row 

void x264_predict_4x4_h_c( pixel *src )
{
    SRC_X4(0,0) = PIXEL_SPLAT_X4( SRC(-1,0) );
    SRC_X4(0,1) = PIXEL_SPLAT_X4( SRC(-1,1) );
    SRC_X4(0,2) = PIXEL_SPLAT_X4( SRC(-1,2) );
    SRC_X4(0,3) = PIXEL_SPLAT_X4( SRC(-1,3) );

}


// mode 0 (Vertical), take top row and copy each of 4x4 block

void x264_predict_4x4_v_c( pixel *src )
{
    PREDICT_4x4_DC(SRC_X4(0,-1));
}

// take left pixel to l0, l1 etc
#define PREDICT_4x4_LOAD_LEFT\
    int l0 = SRC(-1,0);\
    int l1 = SRC(-1,1);\
    int l2 = SRC(-1,2);\
    UNUSED int l3 = SRC(-1,3);

// take top pixel to t0, t1 etc
#define PREDICT_4x4_LOAD_TOP\
    int t0 = SRC(0,-1);\
    int t1 = SRC(1,-1);\
    int t2 = SRC(2,-1);\
    UNUSED int t3 = SRC(3,-1);

// take top right to t4, t5 etc
#define PREDICT_4x4_LOAD_TOP_RIGHT\
    int t4 = SRC(4,-1);\
    int t5 = SRC(5,-1);\
    int t6 = SRC(6,-1);\
    UNUSED int t7 = SRC(7,-1);

// mean of a, b, then round

#define F1(a,b)   (((a)+(b)+1)>>1)

// weighted mean of a, b, c, then round

#define F2(a,b,c) (((a)+2*(b)+(c)+2)>>2)

// mode 3 (Dialog Down Left)
static void x264_predict_4x4_ddl_c( pixel *src )
{
    PREDICT_4x4_LOAD_TOP
    PREDICT_4x4_LOAD_TOP_RIGHT
    SRC(0,0)= F2(t0,t1,t2);   // src[0][0] =  Round(t0, t1, t2's weighted mean)
    SRC(1,0)=SRC(0,1)= F2(t1,t2,t3); // src[0][1] = src[1][0] = Round(t1, t2, t3's weighted mean)
    SRC(2,0)=SRC(1,1)=SRC(0,2)= F2(t2,t3,t4);
    SRC(3,0)=SRC(2,1)=SRC(1,2)=SRC(0,3)= F2(t3,t4,t5);
    SRC(3,1)=SRC(2,2)=SRC(1,3)= F2(t4,t5,t6);
    SRC(3,2)=SRC(2,3)= F2(t5,t6,t7);
    SRC(3,3)= F2(t6,t7,t7);

}


// mode 4 (Dialog Down Right) 

static void x264_predict_4x4_ddr_c( pixel *src )
{
    int lt = SRC(-1,-1);
    PREDICT_4x4_LOAD_LEFT
    PREDICT_4x4_LOAD_TOP
    SRC(3,0)= F2(t3,t2,t1);   // assign src[0][3]
    SRC(2,0)=SRC(3,1)= F2(t2,t1,t0);   // assign src[0][2], src[1][3
    SRC(1,0)=SRC(2,1)=SRC(3,2)= F2(t1,t0,lt);  
    SRC(0,0)=SRC(1,1)=SRC(2,2)=SRC(3,3)= F2(t0,lt,l0);
    SRC(0,1)=SRC(1,2)=SRC(2,3)= F2(lt,l0,l1);
    SRC(0,2)=SRC(1,3)= F2(l0,l1,l2);
    SRC(0,3)= F2(l1,l2,l3);
}

// mode 5 (Vertical Right)
static void x264_predict_4x4_vr_c( pixel *src )
{
    int lt = SRC(-1,-1);
    PREDICT_4x4_LOAD_LEFT
    PREDICT_4x4_LOAD_TOP
    SRC(0,3)= F2(l2,l1,l0);  // assign src[3][0],  weighted mean as l1 center rather than dialog l2 as center
    SRC(0,2)= F2(l1,l0,lt);    // assign src[2][0], weighted mean as l0 center.
    SRC(0,1)=SRC(1,3)= F2(l0,lt,t0);  // assign src[1][0], src[3][1], weigthed mean as lt as center
    SRC(0,0)=SRC(1,2)= F1(lt,t0);      // assign src[0][0], src[2][1], mean of lt, t0
    SRC(1,1)=SRC(2,3)= F2(lt,t0,t1);  //  assign src[1][1], src[3][2], t0 as center
    SRC(1,0)=SRC(2,2)= F1(t0,t1);    // src[0][1], src[2][2];  mean of t0, t1
    SRC(2,1)=SRC(3,3)= F2(t0,t1,t2); // src[1][2], src[3][3]; t1 as center
    SRC(2,0)=SRC(3,2)= F1(t1,t2);     // src[0][2], src[2][3]; mean of t1, t2
    SRC(3,1)= F2(t1,t2,t3);                   //  src[1][3]; t2 as center
    SRC(3,0)= F1(t2,t3);                       //  src[0][3];  mean of t2, t3
}

// mode 6 (Horizon Down)
static void x264_predict_4x4_hd_c( pixel *src )
{
    int lt= SRC(-1,-1);
    PREDICT_4x4_LOAD_LEFT
    PREDICT_4x4_LOAD_TOP
    SRC(0,3)= F1(l2,l3);    // src[3][0]; mean of l2, l3
    SRC(1,3)= F2(l1,l2,l3);   // src[3][1]; as l2 center
    SRC(0,2)=SRC(2,3)= F1(l1,l2);   // src[2][0], src[3][2]; mean of l1, l2
    SRC(1,2)=SRC(3,3)= F2(l0,l1,l2);  // src[2][1], src[3][3]; as l1 as center
    SRC(0,1)=SRC(2,2)= F1(l0,l1);      // src[1][0], src[2][2]; mean of l0, l1
    SRC(1,1)=SRC(3,2)= F2(lt,l0,l1);   // src[1][1], src[2][3]; as l0 as center
    SRC(0,0)=SRC(2,1)= F1(lt,l0);       // src[0][0], src[1][2]; mean of lt, l0
    SRC(1,0)=SRC(3,1)= F2(t0,lt,l0);   // src[0][1], src[1][3]; as lt center
    SRC(2,0)= F2(t1,t0,lt);                      // src[0][2];  as t0 center
    SRC(3,0)= F2(t2,t1,t0);                     // src[0][3];  as t1 center
}

// mode 7 (Vertical Left)
static void x264_predict_4x4_vl_c( pixel *src )
{
    PREDICT_4x4_LOAD_TOP
    PREDICT_4x4_LOAD_TOP_RIGHT
    SRC(0,0)= F1(t0,t1);       // src[0][0]; mean t0, t1
    SRC(0,1)= F2(t0,t1,t2);      // src[1][0]; as t1 center
    SRC(1,0)=SRC(0,2)= F1(t1,t2);     // src[0][1], src[2][0];  mean of t1, t2
    SRC(1,1)=SRC(0,3)= F2(t1,t2,t3);   // src[1][1], src[3][0]; as t2 center
    SRC(2,0)=SRC(1,2)= F1(t2,t3);      // src[0][2], src[2][1]; mean of t2, t3
    SRC(2,1)=SRC(1,3)= F2(t2,t3,t4);    // src[1][2], src[3][1]; as t3 center
    SRC(3,0)=SRC(2,2)= F1(t3,t4);        // src[0][3], src[2][2]; mean of t3, t4
    SRC(3,1)=SRC(2,3)= F2(t3,t4,t5);    // src[1][3], src[3][2]; t4 as center
    SRC(3,2)= F1(t4,t5);                            // src[2][3]; mean of t4, t5
    SRC(3,3)= F2(t4,t5,t6);                        // src[3][3]; as t5 center
}

// mode 8 (horizon up)
static void x264_predict_4x4_hu_c( pixel *src )
{
    PREDICT_4x4_LOAD_LEFT
    SRC(0,0)= F1(l0,l1);                //  src[0][0]; mean of l0, l1
    SRC(1,0)= F2(l0,l1,l2);            //  src[0][1]; as l1 center
    SRC(2,0)=SRC(0,1)= F1(l1,l2);    // src[0][2], src[1][0]; mean of l1, l2
    SRC(3,0)=SRC(1,1)= F2(l1,l2,l3);    // src[0][3], src[1][1]; l2 as center
    SRC(2,1)=SRC(0,2)= F1(l2,l3);        // src[1][2], src[2][0]; mean of l2, l3
    SRC(3,1)=SRC(1,2)= F2(l2,l3,l3);   // src[1][3], src[2][1]; as l3 center
    SRC(3,2)=SRC(1,3)=SRC(0,3)=
    SRC(2,2)=SRC(2,3)=SRC(3,3)= l3;      // others = l3
}


/****************************************************************************
 * 8x8 prediction for intra luma block
 ****************************************************************************/


#define PL(y) \
    edge[14-y] = F2(SRC(-1,y-1), SRC(-1,y), SRC(-1,y+1));
#define PT(x) \
    edge[16+x] = F2(SRC(x-1,-1), SRC(x,-1), SRC(x+1,-1));


static void x264_predict_8x8_filter_c( pixel *src, pixel edge[36], int i_neighbor, int i_filters )
{
    /* edge[7..14] = l7..l0
     * edge[15] = lt
     * edge[16..31] = t0 .. t15
     * edge[32] = t15 */


    int have_lt = i_neighbor & MB_TOPLEFT;
    if( i_filters & MB_LEFT )
    {
        edge[15] = (SRC(0,-1) + 2*SRC(-1,-1) + SRC(-1,0) + 2) >> 2;
        edge[14] = ((have_lt ? SRC(-1,-1) : SRC(-1,0))
                 + 2*SRC(-1,0) + SRC(-1,1) + 2) >> 2;
        PL(1) PL(2) PL(3) PL(4) PL(5) PL(6)
        edge[6] =
        edge[7] = (SRC(-1,6) + 3*SRC(-1,7) + 2) >> 2;
    }


    if( i_filters & MB_TOP )
    {
        int have_tr = i_neighbor & MB_TOPRIGHT;
        edge[16] = ((have_lt ? SRC(-1,-1) : SRC(0,-1))
                 + 2*SRC(0,-1) + SRC(1,-1) + 2) >> 2;
        PT(1) PT(2) PT(3) PT(4) PT(5) PT(6)
        edge[23] = (SRC(6,-1) + 2*SRC(7,-1)
                 + (have_tr ? SRC(8,-1) : SRC(7,-1)) + 2) >> 2;


        if( i_filters & MB_TOPRIGHT )
        {
            if( have_tr )
            {
                PT(8) PT(9) PT(10) PT(11) PT(12) PT(13) PT(14)
                edge[31] =
                edge[32] = (SRC(14,-1) + 3*SRC(15,-1) + 2) >> 2;
            }
            else
            {
                MPIXEL_X4( edge+24 ) = PIXEL_SPLAT_X4( SRC(7,-1) );
                MPIXEL_X4( edge+28 ) = PIXEL_SPLAT_X4( SRC(7,-1) );
                edge[32] = SRC(7,-1);
            }
        }
    }
}


#undef PL
#undef PT


#define PL(y) \
    UNUSED int l##y = edge[14-y];
#define PT(x) \
    UNUSED int t##x = edge[16+x];
#define PREDICT_8x8_LOAD_TOPLEFT \
    int lt = edge[15];
#define PREDICT_8x8_LOAD_LEFT \
    PL(0) PL(1) PL(2) PL(3) PL(4) PL(5) PL(6) PL(7)
#define PREDICT_8x8_LOAD_TOP \
    PT(0) PT(1) PT(2) PT(3) PT(4) PT(5) PT(6) PT(7)
#define PREDICT_8x8_LOAD_TOPRIGHT \
    PT(8) PT(9) PT(10) PT(11) PT(12) PT(13) PT(14) PT(15)


#define PREDICT_8x8_DC(v) \
    for( int y = 0; y < 8; y++ ) { \
        MPIXEL_X4( src+0 ) = v; \
        MPIXEL_X4( src+4 ) = v; \
        src += FDEC_STRIDE; \
    }


static void x264_predict_8x8_dc_128_c( pixel *src, pixel edge[36] )
{
    PREDICT_8x8_DC( PIXEL_SPLAT_X4( 1 << (BIT_DEPTH-1) ) );
}
static void x264_predict_8x8_dc_left_c( pixel *src, pixel edge[36] )
{

//  load l0, ..., l7 from edge
    PREDICT_8x8_LOAD_LEFT

//  取平均值v, 将v延展为四字节， 每个字节都为v, 将延展结果赋值给dc
    pixel4 dc = PIXEL_SPLAT_X4( (l0+l1+l2+l3+l4+l5+l6+l7+4) >> 3 );

// 将dc赋值给8行中的每一行的两个4字节字， 这样8x8块的每一字节都将是v
    PREDICT_8x8_DC( dc );
}
static void x264_predict_8x8_dc_top_c( pixel *src, pixel edge[36] )
{
    PREDICT_8x8_LOAD_TOP
    pixel4 dc = PIXEL_SPLAT_X4( (t0+t1+t2+t3+t4+t5+t6+t7+4) >> 3 );
    PREDICT_8x8_DC( dc );
}
void x264_predict_8x8_dc_c( pixel *src, pixel edge[36] )
{
    PREDICT_8x8_LOAD_LEFT
    PREDICT_8x8_LOAD_TOP
    pixel4 dc = PIXEL_SPLAT_X4( (l0+l1+l2+l3+l4+l5+l6+l7+t0+t1+t2+t3+t4+t5+t6+t7+8) >> 4 );
    PREDICT_8x8_DC( dc );
}
void x264_predict_8x8_h_c( pixel *src, pixel edge[36] )
{
    PREDICT_8x8_LOAD_LEFT
#define ROW(y) MPIXEL_X4( src+y*FDEC_STRIDE+0 ) =\
               MPIXEL_X4( src+y*FDEC_STRIDE+4 ) = PIXEL_SPLAT_X4( l##y );
    ROW(0); ROW(1); ROW(2); ROW(3); ROW(4); ROW(5); ROW(6); ROW(7);
#undef ROW
}
void x264_predict_8x8_v_c( pixel *src, pixel edge[36] )
{
    pixel4 top[2] = { MPIXEL_X4( edge+16 ),
                      MPIXEL_X4( edge+20 ) };
    for( int y = 0; y < 8; y++ )
    {
        MPIXEL_X4( src+y*FDEC_STRIDE+0 ) = top[0];
        MPIXEL_X4( src+y*FDEC_STRIDE+4 ) = top[1];
    }
}
static void x264_predict_8x8_ddl_c( pixel *src, pixel edge[36] )
{
    PREDICT_8x8_LOAD_TOP
    PREDICT_8x8_LOAD_TOPRIGHT
    SRC(0,0)= F2(t0,t1,t2);
    SRC(0,1)=SRC(1,0)= F2(t1,t2,t3);
    SRC(0,2)=SRC(1,1)=SRC(2,0)= F2(t2,t3,t4);
    SRC(0,3)=SRC(1,2)=SRC(2,1)=SRC(3,0)= F2(t3,t4,t5);
    SRC(0,4)=SRC(1,3)=SRC(2,2)=SRC(3,1)=SRC(4,0)= F2(t4,t5,t6);
    SRC(0,5)=SRC(1,4)=SRC(2,3)=SRC(3,2)=SRC(4,1)=SRC(5,0)= F2(t5,t6,t7);
    SRC(0,6)=SRC(1,5)=SRC(2,4)=SRC(3,3)=SRC(4,2)=SRC(5,1)=SRC(6,0)= F2(t6,t7,t8);
    SRC(0,7)=SRC(1,6)=SRC(2,5)=SRC(3,4)=SRC(4,3)=SRC(5,2)=SRC(6,1)=SRC(7,0)= F2(t7,t8,t9);
    SRC(1,7)=SRC(2,6)=SRC(3,5)=SRC(4,4)=SRC(5,3)=SRC(6,2)=SRC(7,1)= F2(t8,t9,t10);
    SRC(2,7)=SRC(3,6)=SRC(4,5)=SRC(5,4)=SRC(6,3)=SRC(7,2)= F2(t9,t10,t11);
    SRC(3,7)=SRC(4,6)=SRC(5,5)=SRC(6,4)=SRC(7,3)= F2(t10,t11,t12);
    SRC(4,7)=SRC(5,6)=SRC(6,5)=SRC(7,4)= F2(t11,t12,t13);
    SRC(5,7)=SRC(6,6)=SRC(7,5)= F2(t12,t13,t14);
    SRC(6,7)=SRC(7,6)= F2(t13,t14,t15);
    SRC(7,7)= F2(t14,t15,t15);
}
static void x264_predict_8x8_ddr_c( pixel *src, pixel edge[36] )
{
    PREDICT_8x8_LOAD_TOP
    PREDICT_8x8_LOAD_LEFT
    PREDICT_8x8_LOAD_TOPLEFT
    SRC(0,7)= F2(l7,l6,l5);
    SRC(0,6)=SRC(1,7)= F2(l6,l5,l4);
    SRC(0,5)=SRC(1,6)=SRC(2,7)= F2(l5,l4,l3);
    SRC(0,4)=SRC(1,5)=SRC(2,6)=SRC(3,7)= F2(l4,l3,l2);
    SRC(0,3)=SRC(1,4)=SRC(2,5)=SRC(3,6)=SRC(4,7)= F2(l3,l2,l1);
    SRC(0,2)=SRC(1,3)=SRC(2,4)=SRC(3,5)=SRC(4,6)=SRC(5,7)= F2(l2,l1,l0);
    SRC(0,1)=SRC(1,2)=SRC(2,3)=SRC(3,4)=SRC(4,5)=SRC(5,6)=SRC(6,7)= F2(l1,l0,lt);
    SRC(0,0)=SRC(1,1)=SRC(2,2)=SRC(3,3)=SRC(4,4)=SRC(5,5)=SRC(6,6)=SRC(7,7)= F2(l0,lt,t0);
    SRC(1,0)=SRC(2,1)=SRC(3,2)=SRC(4,3)=SRC(5,4)=SRC(6,5)=SRC(7,6)= F2(lt,t0,t1);
    SRC(2,0)=SRC(3,1)=SRC(4,2)=SRC(5,3)=SRC(6,4)=SRC(7,5)= F2(t0,t1,t2);
    SRC(3,0)=SRC(4,1)=SRC(5,2)=SRC(6,3)=SRC(7,4)= F2(t1,t2,t3);
    SRC(4,0)=SRC(5,1)=SRC(6,2)=SRC(7,3)= F2(t2,t3,t4);
    SRC(5,0)=SRC(6,1)=SRC(7,2)= F2(t3,t4,t5);
    SRC(6,0)=SRC(7,1)= F2(t4,t5,t6);
    SRC(7,0)= F2(t5,t6,t7);


}
static void x264_predict_8x8_vr_c( pixel *src, pixel edge[36] )
{
    PREDICT_8x8_LOAD_TOP
    PREDICT_8x8_LOAD_LEFT
    PREDICT_8x8_LOAD_TOPLEFT
    SRC(0,6)= F2(l5,l4,l3);
    SRC(0,7)= F2(l6,l5,l4);
    SRC(0,4)=SRC(1,6)= F2(l3,l2,l1);
    SRC(0,5)=SRC(1,7)= F2(l4,l3,l2);
    SRC(0,2)=SRC(1,4)=SRC(2,6)= F2(l1,l0,lt);
    SRC(0,3)=SRC(1,5)=SRC(2,7)= F2(l2,l1,l0);
    SRC(0,1)=SRC(1,3)=SRC(2,5)=SRC(3,7)= F2(l0,lt,t0);
    SRC(0,0)=SRC(1,2)=SRC(2,4)=SRC(3,6)= F1(lt,t0);
    SRC(1,1)=SRC(2,3)=SRC(3,5)=SRC(4,7)= F2(lt,t0,t1);
    SRC(1,0)=SRC(2,2)=SRC(3,4)=SRC(4,6)= F1(t0,t1);
    SRC(2,1)=SRC(3,3)=SRC(4,5)=SRC(5,7)= F2(t0,t1,t2);
    SRC(2,0)=SRC(3,2)=SRC(4,4)=SRC(5,6)= F1(t1,t2);
    SRC(3,1)=SRC(4,3)=SRC(5,5)=SRC(6,7)= F2(t1,t2,t3);
    SRC(3,0)=SRC(4,2)=SRC(5,4)=SRC(6,6)= F1(t2,t3);
    SRC(4,1)=SRC(5,3)=SRC(6,5)=SRC(7,7)= F2(t2,t3,t4);
    SRC(4,0)=SRC(5,2)=SRC(6,4)=SRC(7,6)= F1(t3,t4);
    SRC(5,1)=SRC(6,3)=SRC(7,5)= F2(t3,t4,t5);
    SRC(5,0)=SRC(6,2)=SRC(7,4)= F1(t4,t5);
    SRC(6,1)=SRC(7,3)= F2(t4,t5,t6);
    SRC(6,0)=SRC(7,2)= F1(t5,t6);
    SRC(7,1)= F2(t5,t6,t7);
    SRC(7,0)= F1(t6,t7);
}
static void x264_predict_8x8_hd_c( pixel *src, pixel edge[36] )
{
    PREDICT_8x8_LOAD_TOP
    PREDICT_8x8_LOAD_LEFT
    PREDICT_8x8_LOAD_TOPLEFT
    int p1 = pack_pixel_1to2(F1(l6,l7), F2(l5,l6,l7));
    int p2 = pack_pixel_1to2(F1(l5,l6), F2(l4,l5,l6));
    int p3 = pack_pixel_1to2(F1(l4,l5), F2(l3,l4,l5));
    int p4 = pack_pixel_1to2(F1(l3,l4), F2(l2,l3,l4));
    int p5 = pack_pixel_1to2(F1(l2,l3), F2(l1,l2,l3));
    int p6 = pack_pixel_1to2(F1(l1,l2), F2(l0,l1,l2));
    int p7 = pack_pixel_1to2(F1(l0,l1), F2(lt,l0,l1));
    int p8 = pack_pixel_1to2(F1(lt,l0), F2(l0,lt,t0));
    int p9 = pack_pixel_1to2(F2(t1,t0,lt), F2(t2,t1,t0));
    int p10 = pack_pixel_1to2(F2(t3,t2,t1), F2(t4,t3,t2));
    int p11 = pack_pixel_1to2(F2(t5,t4,t3), F2(t6,t5,t4));
    SRC_X4(0,7)= pack_pixel_2to4(p1,p2);
    SRC_X4(0,6)= pack_pixel_2to4(p2,p3);
    SRC_X4(4,7)=SRC_X4(0,5)= pack_pixel_2to4(p3,p4);
    SRC_X4(4,6)=SRC_X4(0,4)= pack_pixel_2to4(p4,p5);
    SRC_X4(4,5)=SRC_X4(0,3)= pack_pixel_2to4(p5,p6);
    SRC_X4(4,4)=SRC_X4(0,2)= pack_pixel_2to4(p6,p7);
    SRC_X4(4,3)=SRC_X4(0,1)= pack_pixel_2to4(p7,p8);
    SRC_X4(4,2)=SRC_X4(0,0)= pack_pixel_2to4(p8,p9);
    SRC_X4(4,1)= pack_pixel_2to4(p9,p10);
    SRC_X4(4,0)= pack_pixel_2to4(p10,p11);
}
static void x264_predict_8x8_vl_c( pixel *src, pixel edge[36] )
{
    PREDICT_8x8_LOAD_TOP
    PREDICT_8x8_LOAD_TOPRIGHT
    SRC(0,0)= F1(t0,t1);
    SRC(0,1)= F2(t0,t1,t2);
    SRC(0,2)=SRC(1,0)= F1(t1,t2);
    SRC(0,3)=SRC(1,1)= F2(t1,t2,t3);
    SRC(0,4)=SRC(1,2)=SRC(2,0)= F1(t2,t3);
    SRC(0,5)=SRC(1,3)=SRC(2,1)= F2(t2,t3,t4);
    SRC(0,6)=SRC(1,4)=SRC(2,2)=SRC(3,0)= F1(t3,t4);
    SRC(0,7)=SRC(1,5)=SRC(2,3)=SRC(3,1)= F2(t3,t4,t5);
    SRC(1,6)=SRC(2,4)=SRC(3,2)=SRC(4,0)= F1(t4,t5);
    SRC(1,7)=SRC(2,5)=SRC(3,3)=SRC(4,1)= F2(t4,t5,t6);
    SRC(2,6)=SRC(3,4)=SRC(4,2)=SRC(5,0)= F1(t5,t6);
    SRC(2,7)=SRC(3,5)=SRC(4,3)=SRC(5,1)= F2(t5,t6,t7);
    SRC(3,6)=SRC(4,4)=SRC(5,2)=SRC(6,0)= F1(t6,t7);
    SRC(3,7)=SRC(4,5)=SRC(5,3)=SRC(6,1)= F2(t6,t7,t8);
    SRC(4,6)=SRC(5,4)=SRC(6,2)=SRC(7,0)= F1(t7,t8);
    SRC(4,7)=SRC(5,5)=SRC(6,3)=SRC(7,1)= F2(t7,t8,t9);
    SRC(5,6)=SRC(6,4)=SRC(7,2)= F1(t8,t9);
    SRC(5,7)=SRC(6,5)=SRC(7,3)= F2(t8,t9,t10);
    SRC(6,6)=SRC(7,4)= F1(t9,t10);
    SRC(6,7)=SRC(7,5)= F2(t9,t10,t11);
    SRC(7,6)= F1(t10,t11);
    SRC(7,7)= F2(t10,t11,t12);
}
static void x264_predict_8x8_hu_c( pixel *src, pixel edge[36] )
{
    PREDICT_8x8_LOAD_LEFT
    int p1 = pack_pixel_1to2(F1(l0,l1), F2(l0,l1,l2));
    int p2 = pack_pixel_1to2(F1(l1,l2), F2(l1,l2,l3));
    int p3 = pack_pixel_1to2(F1(l2,l3), F2(l2,l3,l4));
    int p4 = pack_pixel_1to2(F1(l3,l4), F2(l3,l4,l5));
    int p5 = pack_pixel_1to2(F1(l4,l5), F2(l4,l5,l6));
    int p6 = pack_pixel_1to2(F1(l5,l6), F2(l5,l6,l7));
    int p7 = pack_pixel_1to2(F1(l6,l7), F2(l6,l7,l7));
    int p8 = pack_pixel_1to2(l7,l7);
    SRC_X4(0,0)= pack_pixel_2to4(p1,p2);
    SRC_X4(0,1)= pack_pixel_2to4(p2,p3);
    SRC_X4(4,0)=SRC_X4(0,2)= pack_pixel_2to4(p3,p4);
    SRC_X4(4,1)=SRC_X4(0,3)= pack_pixel_2to4(p4,p5);
    SRC_X4(4,2)=SRC_X4(0,4)= pack_pixel_2to4(p5,p6);
    SRC_X4(4,3)=SRC_X4(0,5)= pack_pixel_2to4(p6,p7);
    SRC_X4(4,4)=SRC_X4(0,6)= pack_pixel_2to4(p7,p8);
    SRC_X4(4,5)=SRC_X4(4,6)= SRC_X4(0,7) = SRC_X4(4,7) = pack_pixel_2to4(p8,p8);
}


/****************************************************************************
 * Exported functions:
 ****************************************************************************/
void x264_predict_16x16_init( int cpu, x264_predict_t pf[7] )
{
    pf[I_PRED_16x16_V ]     = x264_predict_16x16_v_c;
    pf[I_PRED_16x16_H ]     = x264_predict_16x16_h_c;
    pf[I_PRED_16x16_DC]     = x264_predict_16x16_dc_c;
    pf[I_PRED_16x16_P ]     = x264_predict_16x16_p_c;
    pf[I_PRED_16x16_DC_LEFT]= x264_predict_16x16_dc_left_c;
    pf[I_PRED_16x16_DC_TOP ]= x264_predict_16x16_dc_top_c;
    pf[I_PRED_16x16_DC_128 ]= x264_predict_16x16_dc_128_c;


#if HAVE_MMX
    x264_predict_16x16_init_mmx( cpu, pf );
#endif


#if HAVE_ALTIVEC
    if( cpu&X264_CPU_ALTIVEC )
        x264_predict_16x16_init_altivec( pf );
#endif


#if HAVE_ARMV6
    x264_predict_16x16_init_arm( cpu, pf );
#endif
}


void x264_predict_8x8c_init( int cpu, x264_predict_t pf[7] )
{
    pf[I_PRED_CHROMA_V ]     = x264_predict_8x8c_v_c;
    pf[I_PRED_CHROMA_H ]     = x264_predict_8x8c_h_c;
    pf[I_PRED_CHROMA_DC]     = x264_predict_8x8c_dc_c;
    pf[I_PRED_CHROMA_P ]     = x264_predict_8x8c_p_c;
    pf[I_PRED_CHROMA_DC_LEFT]= x264_predict_8x8c_dc_left_c;
    pf[I_PRED_CHROMA_DC_TOP ]= x264_predict_8x8c_dc_top_c;
    pf[I_PRED_CHROMA_DC_128 ]= x264_predict_8x8c_dc_128_c;


#if HAVE_MMX
    x264_predict_8x8c_init_mmx( cpu, pf );
#endif


#if HAVE_ALTIVEC
    if( cpu&X264_CPU_ALTIVEC )
        x264_predict_8x8c_init_altivec( pf );
#endif


#if HAVE_ARMV6
    x264_predict_8x8c_init_arm( cpu, pf );
#endif
}


void x264_predict_8x16c_init( int cpu, x264_predict_t pf[7] )
{
    pf[I_PRED_CHROMA_V ]     = x264_predict_8x16c_v_c;
    pf[I_PRED_CHROMA_H ]     = x264_predict_8x16c_h_c;
    pf[I_PRED_CHROMA_DC]     = x264_predict_8x16c_dc_c;
    pf[I_PRED_CHROMA_P ]     = x264_predict_8x16c_p_c;
    pf[I_PRED_CHROMA_DC_LEFT]= x264_predict_8x16c_dc_left_c;
    pf[I_PRED_CHROMA_DC_TOP ]= x264_predict_8x16c_dc_top_c;
    pf[I_PRED_CHROMA_DC_128 ]= x264_predict_8x16c_dc_128_c;


#if HAVE_MMX
    x264_predict_8x16c_init_mmx( cpu, pf );
#endif
}


void x264_predict_8x8_init( int cpu, x264_predict8x8_t pf[12], x264_predict_8x8_filter_t *predict_filter )
{
    pf[I_PRED_8x8_V]      = x264_predict_8x8_v_c;
    pf[I_PRED_8x8_H]      = x264_predict_8x8_h_c;
    pf[I_PRED_8x8_DC]     = x264_predict_8x8_dc_c;
    pf[I_PRED_8x8_DDL]    = x264_predict_8x8_ddl_c;
    pf[I_PRED_8x8_DDR]    = x264_predict_8x8_ddr_c;
    pf[I_PRED_8x8_VR]     = x264_predict_8x8_vr_c;
    pf[I_PRED_8x8_HD]     = x264_predict_8x8_hd_c;
    pf[I_PRED_8x8_VL]     = x264_predict_8x8_vl_c;
    pf[I_PRED_8x8_HU]     = x264_predict_8x8_hu_c;
    pf[I_PRED_8x8_DC_LEFT]= x264_predict_8x8_dc_left_c;
    pf[I_PRED_8x8_DC_TOP] = x264_predict_8x8_dc_top_c;
    pf[I_PRED_8x8_DC_128] = x264_predict_8x8_dc_128_c;
    *predict_filter       = x264_predict_8x8_filter_c;


#if HAVE_MMX
    x264_predict_8x8_init_mmx( cpu, pf, predict_filter );
#endif


#if HAVE_ARMV6
    x264_predict_8x8_init_arm( cpu, pf, predict_filter );
#endif
}


void x264_predict_4x4_init( int cpu, x264_predict_t pf[12] )
{
    pf[I_PRED_4x4_V]      = x264_predict_4x4_v_c;
    pf[I_PRED_4x4_H]      = x264_predict_4x4_h_c;
    pf[I_PRED_4x4_DC]     = x264_predict_4x4_dc_c;
    pf[I_PRED_4x4_DDL]    = x264_predict_4x4_ddl_c;
    pf[I_PRED_4x4_DDR]    = x264_predict_4x4_ddr_c;
    pf[I_PRED_4x4_VR]     = x264_predict_4x4_vr_c;
    pf[I_PRED_4x4_HD]     = x264_predict_4x4_hd_c;
    pf[I_PRED_4x4_VL]     = x264_predict_4x4_vl_c;
    pf[I_PRED_4x4_HU]     = x264_predict_4x4_hu_c;
    pf[I_PRED_4x4_DC_LEFT]= x264_predict_4x4_dc_left_c;
    pf[I_PRED_4x4_DC_TOP] = x264_predict_4x4_dc_top_c;
    pf[I_PRED_4x4_DC_128] = x264_predict_4x4_dc_128_c;


#if HAVE_MMX
    x264_predict_4x4_init_mmx( cpu, pf );
#endif


#if HAVE_ARMV6
    x264_predict_4x4_init_arm( cpu, pf );
#endif
}



fanbird2008
关注
专栏目录
重新编译x264解决x264_opencl_load_library 中common/opencl.c:61: undefined reference to `dlopen'
突围
01-18 4610
/home/zhangbin/ffmpeg/ffmpegx86/transcoder/home/lib/libx264.a(opencl.o): In function `x264_opencl_load_library': /home/zhangbin/ffmpeg/ffmpegx86/transcoder/x264build/common/opencl.c:61: undefined ref
x264视频编码中预测模式的实现方法
于军 yuarmy@126.com关注音视频开发
05-09 729
最近学习x264的编码原理，发现
predict.c
wmnmtm的专栏
09-08 203
    predict.c下的内容： -----------------------16x16static void predict_16x16_dc( uint8_t *src )static void predict_16x16_dc_left( uint8_t *src )static void predict_16x16_dc_top( uint8_t *src )static ...
x264_mb_predict_mv_16x16 函数分析
HEVC_CJL的专栏
12-08 3214
void x264_mb_predict_mv_16x16( x264_t *h, int i_list, int i_ref, int16_t mvp[2] ) { int i_refa = h->mb.cache.ref[i_list][X264_SCAN8_0 - 1]; //!< a块参考帧序号 int16_t *mv_a = h->mb.cache.m
x264中16x16帧内预测模式函数分析
HEVC_CJL的专栏
12-05 2737
/**************************************************************************** * 16x16 prediction for intra luma block ****************************************************************************/
x264源码分析之帧内预测
weixin_38189964的博客
06-01 429
x264的帧内预测函数为static void x264_mb_analyse_intra( x264_t *h, x264_mb_analysis_t *res )，下面对该函数进行分析。 /* name: x264_mb_analyse_intra function：对一个16*16的宏块进行帧内预测宏块分析，完成帧内预测模式的选择 1、首先对16*16宏块进行分析，获取最优帧内预测模式 2、然后对4*4宏块进行分析，获取最优帧内预测模式 涉及知识点： 1、16*16帧内预测模式 2、4*4帧内预测模式
logRegres---python.zip_logregres_python 机器学习_逻辑回归_逻辑回归 python
09-23
- **predict_proba()**：如果需要得到每个类别的概率，可以使用`model.predict_proba(X_test)`，它会返回一个二维数组，每一行代表一个样本，列分别对应正类和负类的概率。 6. **模型评估** - **评估指标**：对于...
交叉编译libx264，用于arm上视频采集的h264编码。
烟台 嵌入式产品开发
05-13 5166
先采用软编码调通后，再调用imx6集成的内部h264硬编解码器。 写一个编译脚本： #!/bin/bash ./configure --prefix=/home/shell.albert/libx264arm \ --enable-shared \ --disable-asm \ --host=arm-linux \ --cross-prefix=arm-linux- \ --sy
Kaggle--Predict Future Sales--Python时间序列Gradient Boosting回归
查数菇乐园
05-08 852
相关系数矩阵 GradientBoostingRegressor模型 RMSE检验
TensorFlow2.0学习笔记-3.模型训练
ruoqi23的博客
08-06 2431
3.模型训练 3.1.Keras版本模型训练 •构建模型（顺序模型、函数式模型、子类模型） •模型训练：model.fit() •模型验证：model.evaluate() •模型预测：model.predict() •使用样本加权和类别加权 回调函数 • ModelCheckpoint：定期保存模型。 • EarlyStopping：当训练不再改善验证指标时，停止培训。 • TensorBoard：定期编写可在TensorBoard中可视化的模型日志（更多详细信息，请参见“可视化”...
H.264帧内预测模式详细解释
09-29
H.264帧内预测模式详细解释，希望有所帮助
H.264笔记之三——帧内预测
quanben
08-27 4180
本章讨论的代码主要位于common/predict.c中。x264_macroblock_cache_load函数在每个宏块解码之前初始化某些状态，在x264_slice_write函数的宏块处理循环中被调用。i_mb_xy:   当前宏块的索引i_mb_4x4:  当前宏块中第一个4x4块的索引i_mb_8x8:  当前宏块中第一个8x8块的索引i_top_y:   上方宏块的y索引i_top_
H.264笔记
03-21 2230
H.264标准写得比较繁复，所以考虑在浏览完Whitepaper之后就开始研读X264代码。X264代码风格还是比较清晰简洁的。根据对标准得理解，Picture Order Count在Slice解码的一开始就被提及：I0 B1 B2 P3 B4 B5 P6I0 P3 B1 B2 P6 B4 B5于是I0的POC是0，P3的POC是3，B1是1……为了支持H264复杂的帧存机制
x264源代码简单分析：宏块分析（Analysis）部分-帧内宏块（Intra）
雷霄骅(leixiaohua1020)的专栏
05-22 2万+
本文记录x264的 x264_slice_write()函数中调用的x264_macroblock_analyse()的源代码。x264_macroblock_analyse()对应着x264中的分析模块。分析模块主要完成了下面2个方面的功能： （1）对于帧内宏块，分析帧内预测模式 （2）对于帧间宏块，进行运动估计，分析帧间预测模式
x264源代码简单分析：编码器主干部分-1
热门推荐
雷霄骅(leixiaohua1020)的专栏
05-11 3万+
本文分析x264编码器主干部分的源代码。“主干部分”指的就是libx264中最核心的接口函数——x264_encoder_encode()，以及相关的几个接口函数x264_encoder_open()，x264_encoder_headers()，和x264_encoder_close()。这一部分源代码比较复杂，现在看了半天依然感觉很多地方不太清晰，暂且把已经理解的地方整理出来，以后再慢慢补充还不太清晰的地方。
x264源代码简单分析：编码器主干部分-2
雷霄骅(leixiaohua1020)的专栏
05-14 2万+
本文继续记录x264编码器主干部分的源代码。上一篇文章记录x264_encoder_open()，x264_encoder_headers()，和x264_encoder_close()这三个函数，本文记录x264_encoder_encode()函数。
X264性能优化
zhouyongku的专栏
10-27 1276
一、X264性能分析 测试环境 测试环境：Intel Pentium4 3.00GHz (双核cpu),开启超线程 内存: DDR 1.00G 操作系统: Windows sever 2003 Enterprise Edition 分析软件: Intel(R) VTune(TM) Performance Analyzer 8.0(评估版lic) 编译软件: VC71+nasm0.98 Bus Speed: 800MHz 测试程序: X264 20060506 编码器 1、Debug版本 编码参...
x264阅读记录-1
weixin_34250434的博客
07-19 132
x264阅读记录-1 采用x264版本是x264-snapshot-20060316-2245。 1. main函数 x264的main函数位于x264.c中，下面是main函数调用情况： （1）_setmode函数和_fileno函数 这两个函数是微软提供的两个库函数。 _setmode函数位于io.h文件中，主要作用是设置特定模式匹配的文件。http://ms...
x264代码剖析（十二）：核心算法之帧内预测函数x264_mb_analyse_intra()
weixin_34018202的博客
03-14 653
x264代码剖析（十二）：核心算法之帧内预测函数x264_mb_analyse_intra() 在帧内预测模式中，预测块是基于已编码重建块和当前块形成的，编码器通常选取预测块与编码块之间差异最小的预测模式。H.264帧内预测则是参考预测块左方或者上方的已编码块的临近像素点被引入空间域。 对于亮度像素而言，预测块...
用[[[-6.23,-0.25,-5.46,1.59,-4.6,5.36],[4.58,4.34,4.9,8.49,6.35,14.25],[-0.631,-0.755,-0.959,-0.987,-1.083,-0.927]],-0.43],[[[ -0.25,-5.46, 1.59,-4.6,5.36,-0.43],[4.34,4.9,8.49,6.35,14.25,19.76],[-0.755,-0.959,-0.987,-1.083,-0.927,-0.783]],1.43],]类似的数据作为数据集训练飞浆模型并预测一个值
最新发布
06-06
好的，我们可以使用飞桨（PaddlePaddle）框架训练和预测。首先，我们需要导入必要的库和模块： ```python import paddle import numpy as np ``` 接下来，我们可以定义一个简单的全连接神经网络模型： ```python class MyNet(paddle.nn.Layer): def __init__(self): super(MyNet, self).__init__() self.fc1 = paddle.nn.Linear(6, 10) self.fc2 = paddle.nn.Linear(10, 1) def forward(self, x): x = self.fc1(x) x = paddle.nn.functional.relu(x) x = self.fc2(x) return x ``` 定义好模型之后，我们可以加载数据集并进行训练： ```python # 加载数据集 data = np.array([[[-6.23,-0.25,-5.46,1.59,-4.6,5.36],[4.58,4.34,4.9,8.49,6.35,14.25],[-0.631,-0.755,-0.959,-0.987,-1.083,-0.927]],-0.43],[[[ -0.25,-5.46, 1.59,-4.6,5.36,-0.43],[4.34,4.9,8.49,6.35,14.25,19.76],[-0.755,-0.959,-0.987,-1.083,-0.927,-0.783]],1.43]]) x_data = data[:, :-1, :].reshape(-1, 6).astype('float32') y_data = data[:, -1, :].reshape(-1, 1).astype('float32') # 定义数据读取器 def reader(): for i in range(len(x_data)): yield x_data[i], y_data[i] # 定义优化器和损失函数 model = MyNet() opt = paddle.optimizer.Adam(learning_rate=0.01, parameters=model.parameters()) loss_fn = paddle.nn.MSELoss(reduction='mean') # 训练模型 epochs = 1000 for epoch in range(epochs): for x, y in paddle.batch(reader(), batch_size=1): x = paddle.to_tensor(x) y = paddle.to_tensor(y) y_pred = model(x) loss = loss_fn(y_pred, y) loss.backward() opt.step() opt.clear_grad() if epoch % 100 == 0: print('Epoch {}, loss {}'.format(epoch, loss.numpy())) ``` 训练完成后，我们就可以使用训练好的模型进行预测了： ```python # 使用训练好的模型进行预测 x_test = np.array([[-3.14, 0.25, -1.5, 2.5, -2.8, 3.5]], dtype='float32') model.eval() predict = model(paddle.to_tensor(x_test)) print('Prediction:', predict.numpy()) ``` 请注意，上面的代码只是一个简单示例，实际应用中需要根据具体情况进行适当的修改。
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值