[x264-devel] Re: [PATCH]4x4 intra prediction

Laurent Aimar fenrir at via.ecp.fr
Mon Oct 11 12:45:05 CEST 2004


Hi,

On Fri, Sep 24, 2004, Xiaojie Huang wrote:
> Index: macroblock.c
> ===================================================================
> --- macroblock.c	(revision 48)
> +++ macroblock.c	(working copy)
> @@ -562,16 +562,23 @@
>      {
>          for( i = 0; i < 16; i++ )
>          {
> -            const int i_dst = h->mb.pic.i_stride[0];
> -            uint8_t  *p_dst = &h->mb.pic.p_fdec[0][4 * block_idx_x[i] + 4 * block_idx_y[i] * i_dst];
> -            int      i_mode = h->mb.cache.intra4x4_pred_mode[x264_scan8[i]];
> +            int      i_mode = h->mb.cache.intra4x4_pred_mode[x264_scan8[i]];
> +
> +			/* We don't need x264_mb_encode_i4x4 again since 
> +			 * it has been performed in x264_mb_analyse_intra
> +			 * Xiajie Huang, hworldjj at yahoo.com
> +			 */
> +/*
> +            const int i_dst = h->mb.pic.i_stride[0];
> +            uint8_t  *p_dst = &h->mb.pic.p_fdec[0][4 * block_idx_x[i] + 4 * block_idx_y[i] * i_dst];
> +
> +            / * Do the right prediction * /
> +            h->predict_4x4[i_mode]( p_dst, i_dst );
> +
> +            / * encode one 4x4 block * /
> +            x264_mb_encode_i4x4( h, i, i_qscale );
> +*/
>  
> -            /* Do the right prediction */
> -            h->predict_4x4[i_mode]( p_dst, i_dst );
> -
> -            /* encode one 4x4 block */
> -            x264_mb_encode_i4x4( h, i, i_qscale );
> -
>              /* fix the pred mode value */
>              h->mb.cache.intra4x4_pred_mode[x264_scan8[i]] = x264_mb_pred_mode4x4_fix[i_mode];
>          }

 Thanks for the patch, but I won't apply it (sorry for the delay).

 The main thing is that as x264 is far from stable, I would avoid
to have depedency between what analyse.c does and how we encode the
bitstream. (It will leads to problems for people who wants to try to
change analyse.c).

 But, your patch will interest people who need speed up :)

-- 
fenrir

-- 
This is the x264-devel mailing-list
To unsubscribe, go to: http://developers.videolan.org/lists.html



More information about the x264-devel mailing list