macroblock.c

另一个版本的x264的decoder

        if( !IS_SKIP( h->mb.type[i_mb_t] ) )
        {
            mvc[i][0] = mvr[i_mb_t][0];
            mvc[i][1] = mvr[i_mb_t][1];
            i++;
        }

        if( h->mb.i_mb_x > 0 && !IS_SKIP( h->mb.type[i_mb_t - 1] ) )
        {
            mvc[i][0] = mvr[i_mb_t - 1][0];
            mvc[i][1] = mvr[i_mb_t - 1][1];
            i++;
        }
        if( h->mb.i_mb_x < h->mb.i_mb_stride - 1 && !IS_SKIP( h->mb.type[i_mb_t + 1] ) )
        {
            mvc[i][0] = mvr[i_mb_t + 1][0];
            mvc[i][1] = mvr[i_mb_t + 1][1];
            i++;
        }
    }
    *i_mvc = i;
}
#endif



static inline void x264_mb_mc_01xywh( x264_t *h, int x, int y, int width, int height )
{
    const int i8 = x264_scan8[0]+x+8*y;

    const int i_ref1 = h->mb.cache.ref[1][i8];
    const int mvx1   = h->mb.cache.mv[1][i8][0];//x264_clip3( h->mb.cache.mv[1][i8][0], h->mb.mv_min[0], h->mb.mv_max[0] );
    const int mvy1   = h->mb.cache.mv[1][i8][1];//x264_clip3( h->mb.cache.mv[1][i8][1], h->mb.mv_min[1], h->mb.mv_max[1] );
    DECLARE_ALIGNED( uint8_t, tmp[16*16], 16 );
    int i_mode = x264_size2pixel[height][width];

	x264_mb_mc_0xywh( h, x, y, width, height );
	weight_pred_init( h, 16, i_ref1 );
	h->mc.mc_luma( h->mb.pic.p_fref[1][i_ref1], h->mb.pic.i_stride[0],
                    tmp, 16, mvx1 + 4*4*x, mvy1 + 4*4*y, 4*width, 4*height );
	if( h->sh.pps->b_weighted_bipred )
     {
          unsigned char *L=&(h->mb.pic.p_fdec[0][4*y*h->mb.pic.i_stride[0]+4*x]);
          uint8_t *Cb=&(h->mb.pic.p_fdec[1][2*y*h->mb.pic.i_stride[1]+2*x]);
          uint8_t *Cr=&(h->mb.pic.p_fdec[2][2*y*h->mb.pic.i_stride[2]+2*x]);
		  int i=0;
          int j=0;
          
        for(j=0; j<4*height; j++)
            {
            for(i=0; i<4*width; i++)
                {
                L[i]=clip_uint8((( L[i]*weight_pred_para.w0_L+tmp[i+j*h->mb.pic.i_stride[0]]*weight_pred_para.w1_L+2*weight_pred_para.logWD_L)>>(weight_pred_para.logWD_L+1)+(weight_pred_para.o0_L+weight_pred_para.o1_L+1)>>1));
                }
                L+=h->mb.pic.i_stride[0];
				
            }
        h->mc.mc_chroma( &h->mb.pic.p_fref[1][i_ref1][4][2*y*h->mb.pic.i_stride[1]+2*x], h->mb.pic.i_stride[1],
                          tmp, 16, mvx1, mvy1, 2*width, 2*height );
       
        /* Cb weight pred*/
          for(j=0; j<2*height; j++)
          {
            for(i=0;i<2*width;i++)
            {
            Cb[i]=clip_uint8(((Cb[i]*weight_pred_para.w0_C+tmp[i+j*h->mb.pic.i_stride[0]]*weight_pred_para.w1_C+2*weight_pred_para.logWD_C)>>(weight_pred_para.logWD_C+1)+(weight_pred_para.o0_C+weight_pred_para.o1_C+1)>>1));   
            }
                    Cb+=h->mb.pic.i_stride[1];
            }
        h->mc.mc_chroma( &h->mb.pic.p_fref[1][i_ref1][5][2*y*h->mb.pic.i_stride[2]+2*x], h->mb.pic.i_stride[2],
                          tmp, 16, mvx1, mvy1, 2*width, 2*height );
        
           for(j=0; j<2*height; j++)
          {
            for(i=0;i<2*width;i++)
            {
            Cr[i]=clip_uint8(((Cr[i]*weight_pred_para.w0_C+tmp[i+j*h->mb.pic.i_stride[0]]*weight_pred_para.w1_C+2*weight_pred_para.logWD_C)>>(weight_pred_para.logWD_C+1))+((weight_pred_para.o0_C+weight_pred_para.o1_C+1)>>1));   
            }
                    Cr+=h->mb.pic.i_stride[2];
            }
        }
else
    {
        h->pixf.avg[i_mode]( &h->mb.pic.p_fdec[0][4*y *h->mb.pic.i_stride[0]+4*x], h->mb.pic.i_stride[0], tmp, 16 );

        h->mc.mc_chroma( &h->mb.pic.p_fref[1][i_ref1][4][2*y*h->mb.pic.i_stride[1]+2*x], h->mb.pic.i_stride[1],
                          tmp, 16, mvx1, mvy1, 2*width, 2*height );
        h->pixf.avg[i_mode+3]( &h->mb.pic.p_fdec[1][2*y*h->mb.pic.i_stride[1]+2*x], h->mb.pic.i_stride[1], tmp, 16 );

        h->mc.mc_chroma( &h->mb.pic.p_fref[1][i_ref1][5][2*y*h->mb.pic.i_stride[2]+2*x], h->mb.pic.i_stride[2],
                          tmp, 16, mvx1, mvy1, 2*width, 2*height );
        h->pixf.avg[i_mode+3]( &h->mb.pic.p_fdec[2][2*y*h->mb.pic.i_stride[2]+2*x], h->mb.pic.i_stride[2], tmp, 16 );
    }
  }

static inline void x264_mb_mc_0xywh( x264_t *h, int x, int y, int width, int height )
{
    const int i8 = x264_scan8[0]+x+8*y;
    const int i_ref = h->mb.cache.ref[0][i8];
    const int mvx   = h->mb.cache.mv[0][i8][0];//x264_clip3( h->mb.cache.mv[0][i8][0], h->mb.mv_min[0], h->mb.mv_max[0] );
    const int mvy   = h->mb.cache.mv[0][i8][1];//x264_clip3( h->mb.cache.mv[0][i8][1], h->mb.mv_min[1], h->mb.mv_max[1] );

	weight_pred_init( h,  i_ref, 16 );

    h->mc.mc_luma( h->mb.pic.p_fref[0][i_ref], h->mb.pic.i_stride[0],
                    &h->mb.pic.p_fdec[0][4*y * h->mb.pic.i_stride[0]+4*x],           h->mb.pic.i_stride[0],
                    mvx + 4*4*x, mvy + 4*4*y, 4*width, 4*height );

    h->mc.mc_chroma( &h->mb.pic.p_fref[0][i_ref][4][2*y*h->mb.pic.i_stride[1]+2*x], h->mb.pic.i_stride[1],
                      &h->mb.pic.p_fdec[1][2*y*h->mb.pic.i_stride[1]+2*x],           h->mb.pic.i_stride[1],
                      mvx, mvy, 2*width, 2*height );

    h->mc.mc_chroma( &h->mb.pic.p_fref[0][i_ref][5][2*y*h->mb.pic.i_stride[2]+2*x], h->mb.pic.i_stride[2],
                      &h->mb.pic.p_fdec[2][2*y*h->mb.pic.i_stride[2]+2*x],           h->mb.pic.i_stride[2],
                      mvx, mvy, 2*width, 2*height );
	if((h->sh.pps->b_weighted_pred) && (!h->sh.pps->b_weighted_bipred))
        {
          uint8_t *L=&h->mb.pic.p_fdec[0][4*y * h->mb.pic.i_stride[0]+4*x];
          uint8_t *Cb=&h->mb.pic.p_fdec[1][2*y*h->mb.pic.i_stride[1]+2*x];
          uint8_t *Cr=&h->mb.pic.p_fdec[2][2*y*h->mb.pic.i_stride[2]+2*x];
          int i=0;
          int j=0;
          
          /* Luma weight pred */
         for(j=0; j<4*height;j++)
         {
            for(i=0;i<4*width;i++)
            {
                if(weight_pred_para.logWD_L>1)
                    L[i]=clip_uint8((L[i]*weight_pred_para.w1_L+2*weight_pred_para.logWD_L-1)>>weight_pred_para.logWD_L+weight_pred_para.o1_L);
                else L[i]=clip_uint8((L[i]*weight_pred_para.w1_L+weight_pred_para.o1_L));
                
            }
                    L+=h->mb.pic.i_stride[0];
          }

          /* Cb weight pred*/
          for(j=0; j<2*height; j++)
          {
            for(i=0;i<2*width;i++)
            {
                if(weight_pred_para.logWD_C>1)
                    Cb[i]=clip_uint8((Cb[i]*weight_pred_para.w1_C+2*weight_pred_para.logWD_C-1)>>weight_pred_para.logWD_C+weight_pred_para.o1_C);
                else Cb[i]=clip_uint8((Cb[i]*weight_pred_para.w1_C+weight_pred_para.o1_C));
            }
                    Cb+=h->mb.pic.i_stride[1];
            }
          
          /* Cr weight pred*/
          for(j=0; j<2*height; j++)
          {
            for(i=0;i<2*width;i++)
            {
                if(weight_pred_para.logWD_C>1)
                    Cr[i]=clip_uint8((Cr[i]*weight_pred_para.w1_C+2*weight_pred_para.logWD_C-1)>>weight_pred_para.logWD_C+weight_pred_para.o1_C);
                else Cr[i]=clip_uint8((Cr[i]*weight_pred_para.w1_C+weight_pred_para.o1_C));
            }
                    Cr+=h->mb.pic.i_stride[2];
            }
        }
}


static inline void x264_mb_mc_1xywh( x264_t *h, int x, int y, int width, int height )
{
    const int i8 = x264_scan8[0]+x+8*y;
    const int i_ref = h->mb.cache.ref[1][i8];
    const int mvx   = h->mb.cache.mv[1][i8][0];//x264_clip3( h->mb.cache.mv[1][i8][0], h->mb.mv_min[0], h->mb.mv_max[0] );
    const int mvy   = h->mb.cache.mv[1][i8][1];//x264_clip3( h->mb.cache.mv[1][i8][1], h->mb.mv_min[1], h->mb.mv_max[1] );

	 weight_pred_init( h, 16, i_ref );

    h->mc.mc_luma( h->mb.pic.p_fref[1][i_ref], h->mb.pic.i_stride[0],
                    &h->mb.pic.p_fdec[0][4*y *h->mb.pic.i_stride[0]+4*x],            h->mb.pic.i_stride[0],
                    mvx + 4*4*x, mvy + 4*4*y, 4*width, 4*height );

    h->mc.mc_chroma( &h->mb.pic.p_fref[1][i_ref][4][2*y*h->mb.pic.i_stride[1]+2*x], h->mb.pic.i_stride[1],
                      &h->mb.pic.p_fdec[1][2*y*h->mb.pic.i_stride[1]+2*x],           h->mb.pic.i_stride[1],
                      mvx, mvy, 2*width, 2*height );

    h->mc.mc_chroma( &h->mb.pic.p_fref[1][i_ref][5][2*y*h->mb.pic.i_stride[2]+2*x], h->mb.pic.i_stride[2],
                      &h->mb.pic.p_fdec[2][2*y*h->mb.pic.i_stride[2]+2*x],           h->mb.pic.i_stride[2],
                      mvx, mvy, 2*width, 2*height); 
	if((h->sh.pps->b_weighted_pred) && (!h->sh.pps->b_weighted_bipred))
        {
          uint8_t *L=&h->mb.pic.p_fdec[0][4*y * h->mb.pic.i_stride[0]+4*x];
          uint8_t *Cb=&h->mb.pic.p_fdec[1][2*y*h->mb.pic.i_stride[1]+2*x];
          uint8_t *Cr=&h->mb.pic.p_fdec[2][2*y*h->mb.pic.i_stride[2]+2*x];
          int i=0;
          int j=0;
          /* Luma weight pred */
          for(j=0; j<4*height;j++)
         {
            for(i=0;i<4*width;i++)
            {
                if(weight_pred_para.logWD_L>1)
                    L[i]=clip_uint8(((L[i]*weight_pred_para.w1_L+2*weight_pred_para.logWD_L-1)>>weight_pred_para.logWD_L)+weight_pred_para.o1_L);
                else L[i]=clip_uint8((L[i]*weight_pred_para.w1_L+weight_pred_para.o1_L));
                
            }
                    L+=h->mb.pic.i_stride[0];
          }

          /* Cb weight pred*/
          for(j=0; j<2*height; j++)
          {
            for(i=0;i<2*width;i++)
            {
                if(weight_pred_para.logWD_C>1)
                    Cb[i]=clip_uint8(((Cb[i]*weight_pred_para.w1_C+2*weight_pred_para.logWD_C-1)>>weight_pred_para.logWD_C)+weight_pred_para.o1_C);
                else Cb[i]=clip_uint8((Cb[i]*weight_pred_para.w1_C+weight_pred_para.o1_C));
            }
                    Cb+=h->mb.pic.i_stride[1];
            }
          /* Cr weight pred*/
          for(j=0; j<2*height; j++)
          {
            for(i=0;i<2*width;i++)
            {
                if(weight_pred_para.logWD_C>1)
                    Cr[i]=clip_uint8((Cr[i]*weight_pred_para.w1_C+2*weight_pred_para.logWD_C-1)>>weight_pred_para.logWD_C+weight_pred_para.o1_C);
                else Cr[i]=clip_uint8((Cr[i]*weight_pred_para.w1_C+weight_pred_para.o1_C));
            }
                    Cr+=h->mb.pic.i_stride[2];
            }
        }
}
static void x264_mb_mc_direct8x8( x264_t *h, int x, int y )
{
    const int i8 = x264_scan8[0] + x + 8*y;

    /* FIXME: optimize based on current block size, not global settings? */
    if( h->sps->b_direct8x8_inference )
    {
        if( h->mb.cache.ref[0][i8] >= 0 )
            if( h->mb.cache.ref[1][i8] >= 0 )
                x264_mb_mc_01xywh( h, x, y, 2, 2 );
            else
                x264_mb_mc_0xywh( h, x, y, 2, 2 );
        else
            x264_mb_mc_1xywh( h, x, y, 2, 2 );
    }
    else
    {
        if( h->mb.cache.ref[0][i8] >= 0 )
        {
            if( h->mb.cache.ref[1][i8] >= 0 )
            {
                x264_mb_mc_01xywh( h, x+0, y+0, 1, 1 );
                x264_mb_mc_01xywh( h, x+1, y+0, 1, 1 );
                x264_mb_mc_01xywh( h, x+0, y+1, 1, 1 );
                x264_mb_mc_01xywh( h, x+1, y+1, 1, 1 );
            }
            else
            {
                x264_mb_mc_0xywh( h, x+0, y+0, 1, 1 );
                x264_mb_mc_0xywh( h, x+1, y+0, 1, 1 );
                x264_mb_mc_0xywh( h, x+0, y+1, 1, 1 );
                x264_mb_mc_0xywh( h, x+1, y+1, 1, 1 );
            }
        }
        else
        {
            x264_mb_mc_1xywh( h, x+0, y+0, 1, 1 );
            x264_mb_mc_1xywh( h, x+1, y+0, 1, 1 );
            x264_mb_mc_1xywh( h, x+0, y+1, 1, 1 );
            x264_mb_mc_1xywh( h, x+1, y+1, 1, 1 );
        }
    }
}

void weight_pred_init(x264_t *h, int i_ref0, int i_ref1)
{   
    int refIdxL0=0;
    int refIdxL1=0;
    int i=0;
    
    if((h->sh.i_type==SLICE_TYPE_P && h->sh.pps->b_weighted_pred)
        || (h->sh.i_type==SLICE_TYPE_B && h->sh.pps->b_weighted_bipred))
        {
            
            if(h->sh.b_field_pic=1)
                {
                weight_pred_para.refIdxL0WP=refIdxL0>>1;
                weight_pred_para.refIdxL1WP=refIdxL1>>1;
                }
            else
                {
                weight_pred_para.refIdxL0WP=refIdxL0;
                weight_pred_para.refIdxL1WP=refIdxL1;
                }

            weight_pred_para.logWD_L=h->sh.pred_weight_table.luma_log2_weight_denom;
            weight_pred_para.logWD_C=h->sh.pred_weight_table.chroma_log2_weight_denom;
            
            if(refIdxL0<16)
           {
                weight_pred_para.w0_L=h->sh.pred_weight_table.luma_weight_l0[weight_pred_para.refIdxL0WP];
                weight_pred_para.o0_L=h->sh.pred_weight_table.luma_offset_l0[weight_pred_para.refIdxL0WP];
                for(i=0;i<2;i++)
                {
                weight_pred_para.w0_C=h->sh.pred_weight_table.chroma_weight_l0[weight_pred_para.refIdxL0WP][i];
                weight_pred_para.o0_C=h->sh.pred_weight_table.chroma_offset_l0[weight_pred_para.refIdxL0WP][i];
                }
                
            }
            
            if(refIdxL1<16)
            {
                 weight_pred_para.w1_L=h->sh.pred_weight_table.luma_weight_l1[weight_pred_para.refIdxL1WP];
                 weight_pred_para.o1_L=h->sh.pred_weight_table.luma_offset_l1[weight_pred_para.refIdxL0WP];
                 for(i=0;i<2;i++)
                {
                 weight_pred_para.w1_C=h->sh.pred_weight_table.chroma_weight_l1[weight_pred_para.refIdxL1WP][i];
                 weight_pred_para.o1_C=h->sh.pred_weight_table.chroma_offset_l1[weight_pred_para.refIdxL0WP][i];            
                }
            }
           
		}  
    }

void x264_mb_mc( x264_t *h )
{
    if( h->mb.i_type == P_L0 )
    {
        if( h->mb.i_partition == D_16x16 )
        {
            x264_mb_mc_0xywh( h, 0, 0, 4, 4 );
        }
        else if( h->mb.i_partition == D_16x8 )
        {
            x264_mb_mc_0xywh( h, 0, 0, 4, 2 );
            x264_mb_mc_0xywh( h, 0, 2, 4, 2 );
        }
        else if( h->mb.i_partition == D_8x16 )
        {
            x264_mb_mc_0xywh( h, 0, 0, 2, 4 );
            x264_mb_mc_0xywh( h, 2, 0, 2, 4 );
        }
    }
    else if( h->mb.i_type == P_8x8 || h->mb.i_type == B_8x8 )
    {
        int i;
        for( i = 0; i < 4; i++ )
        {
            const int x = 2*(i%2);
            const int y = 2*(i/2);
            switch( h->mb.i_sub_partition[i] )
            {
                case D_L0_8x8:
                    x264_mb_mc_0xywh( h, x, y, 2, 2 );
                    break;
                case D_L0_8x4:
                    x264_mb_mc_0xywh( h, x, y+0, 2, 1 );
                    x264_mb_mc_0xywh( h, x, y+1, 2, 1 );
                    break;
                case D_L0_4x8:
                    x264_mb_mc_0xywh( h, x+0, y, 1, 2 );
                    x264_mb_mc_0xywh( h, x+1, y, 1, 2 );
                    break;
                case D_L0_4x4:
                    x264_mb_mc_0xywh( h, x+0, y+0, 1, 1 );
                    x264_mb_mc_0xywh( h, x+1, y+0, 1, 1 );
                    x264_mb_mc_0xywh( h, x+0, y+1, 1, 1 );
                    x264_mb_mc_0xywh( h, x+1, y+1, 1, 1 );
                    break;
                case D_L1_8x8:
                    x264_mb_mc_1xywh( h, x, y, 2, 2 );