i420_rgb16.c

video linux conference

            }
        }
        else
        {
            /* 16bpp 5/6/5 */
            for ( i_x = p_vout->render.i_width / 8; i_x--; )
            {
                __asm__( MMX_INIT_16
                         : : "r" (p_y), "r" (p_u), "r" (p_v), "r" (p_buffer) );

                __asm__( ".align 8"
                         MMX_YUV_MUL
                         MMX_YUV_ADD
                         MMX_UNPACK_16
                         : : "r" (p_y), "r" (p_u), "r" (p_v), "r" (p_buffer) );

                p_y += 8;
                p_u += 4;
                p_v += 4;
                p_buffer += 8;
            }
        }
#endif

        /* Here we do some unaligned reads and duplicate conversions, but
         * at least we have all the pixels */
        if( i_rewind )
        {
            p_y -= i_rewind;
            p_u -= i_rewind >> 1;
            p_v -= i_rewind >> 1;
            p_buffer -= i_rewind;
#if defined (MODULE_NAME_IS_i420_rgb)
            CONVERT_YUV_PIXEL(2);  CONVERT_Y_PIXEL(2);
            CONVERT_YUV_PIXEL(2);  CONVERT_Y_PIXEL(2);
            CONVERT_YUV_PIXEL(2);  CONVERT_Y_PIXEL(2);
            CONVERT_YUV_PIXEL(2);  CONVERT_Y_PIXEL(2);
#elif defined (MODULE_NAME_IS_i420_rgb_mmx)
            __asm__( MMX_INIT_16
                     : : "r" (p_y), "r" (p_u), "r" (p_v), "r" (p_buffer) );

            if( p_vout->output.i_rmask == 0x7c00 )
            {
                /* 15bpp 5/5/5 */
                __asm__( ".align 8"
                         MMX_YUV_MUL
                         MMX_YUV_ADD
                         MMX_UNPACK_15
                         : : "r" (p_y), "r" (p_u), "r" (p_v), "r" (p_buffer) );
            }
            else
            {
                /* 16bpp 5/6/5 */
                __asm__( ".align 8"
                         MMX_YUV_MUL
                         MMX_YUV_ADD
                         MMX_UNPACK_16
                         : : "r" (p_y), "r" (p_u), "r" (p_v), "r" (p_buffer) );
            }

            p_y += 8;
            p_u += 4;
            p_v += 4;
            p_buffer += 8;
#endif
        }
        SCALE_WIDTH;
        SCALE_HEIGHT( 420, 2 );

        p_y += i_source_margin;
        if( i_y % 2 )
        {
            p_u += i_source_margin_c;
            p_v += i_source_margin_c;
        }
    }
}

/*****************************************************************************
 * I420_RGB32: color YUV 4:2:0 to RGB 32 bpp
 *****************************************************************************
 * Horizontal alignment needed:
 *  - input: 8 pixels (8 Y bytes, 4 U/V bytes), margins not allowed
 *  - output: 1 pixel (2 bytes), margins allowed
 * Vertical alignment needed:
 *  - input: 2 lines (2 Y lines, 1 U/V line)
 *  - output: 1 line
 *****************************************************************************/
void E_(I420_RGB32)( vout_thread_t *p_vout, picture_t *p_src,
                                            picture_t *p_dest )
{
    /* We got this one from the old arguments */
    uint32_t *p_pic = (uint32_t*)p_dest->p->p_pixels;
    uint8_t  *p_y   = p_src->Y_PIXELS;
    uint8_t  *p_u   = p_src->U_PIXELS;
    uint8_t  *p_v   = p_src->V_PIXELS;

    vlc_bool_t  b_hscale;                         /* horizontal scaling type */
    unsigned int i_vscale;                          /* vertical scaling type */
    unsigned int i_x, i_y;                /* horizontal and vertical indexes */

    int         i_right_margin;
    int         i_rewind;
    int         i_scale_count;                       /* scale modulo counter */
    int         i_chroma_width = p_vout->render.i_width / 2; /* chroma width */
    uint32_t *  p_pic_start;       /* beginning of the current line for copy */
#if defined (MODULE_NAME_IS_i420_rgb)
    int         i_uval, i_vval;                           /* U and V samples */
    int         i_red, i_green, i_blue;          /* U and V modified samples */
    uint32_t *  p_yuv = p_vout->chroma.p_sys->p_rgb32;
    uint32_t *  p_ybase;                     /* Y dependant conversion table */
#endif

    /* Conversion buffer pointer */
    uint32_t *  p_buffer_start = (uint32_t*)p_vout->chroma.p_sys->p_buffer;
    uint32_t *  p_buffer;

    /* Offset array pointer */
    int *       p_offset_start = p_vout->chroma.p_sys->p_offset;
    int *       p_offset;

    const int i_source_margin = p_src->p[0].i_pitch
                                 - p_src->p[0].i_visible_pitch;
    const int i_source_margin_c = p_src->p[1].i_pitch
                                 - p_src->p[1].i_visible_pitch;

    i_right_margin = p_dest->p->i_pitch - p_dest->p->i_visible_pitch;

    if( p_vout->render.i_width & 7 )
    {
        i_rewind = 8 - ( p_vout->render.i_width & 7 );
    }
    else
    {
        i_rewind = 0;
    }

    /* Rule: when a picture of size (x1,y1) with aspect ratio r1 is rendered
     * on a picture of size (x2,y2) with aspect ratio r2, if x1 grows to x1'
     * then y1 grows to y1' = x1' * y2/x2 * r2/r1 */
    SetOffset( p_vout->render.i_width, p_vout->render.i_height,
               p_vout->output.i_width, p_vout->output.i_height,
               &b_hscale, &i_vscale, p_offset_start );

    /*
     * Perform conversion
     */
    i_scale_count = ( i_vscale == 1 ) ?
                    p_vout->output.i_height : p_vout->render.i_height;
    for( i_y = 0; i_y < p_vout->render.i_height; i_y++ )
    {
        p_pic_start = p_pic;
        p_buffer = b_hscale ? p_buffer_start : p_pic;

        for ( i_x = p_vout->render.i_width / 8; i_x--; )
        {
#if defined (MODULE_NAME_IS_i420_rgb)
            CONVERT_YUV_PIXEL(4);  CONVERT_Y_PIXEL(4);
            CONVERT_YUV_PIXEL(4);  CONVERT_Y_PIXEL(4);
            CONVERT_YUV_PIXEL(4);  CONVERT_Y_PIXEL(4);
            CONVERT_YUV_PIXEL(4);  CONVERT_Y_PIXEL(4);
#elif defined (MODULE_NAME_IS_i420_rgb_mmx)
            __asm__( MMX_INIT_32
                     : : "r" (p_y), "r" (p_u), "r" (p_v), "r" (p_buffer) );

            __asm__( ".align 8"
                     MMX_YUV_MUL
                     MMX_YUV_ADD
                     MMX_UNPACK_32
                     : : "r" (p_y), "r" (p_u), "r" (p_v), "r" (p_buffer) );

            p_y += 8;
            p_u += 4;
            p_v += 4;
            p_buffer += 8;
#endif
        }

        /* Here we do some unaligned reads and duplicate conversions, but
         * at least we have all the pixels */
        if( i_rewind )
        {
            p_y -= i_rewind;
            p_u -= i_rewind >> 1;
            p_v -= i_rewind >> 1;
            p_buffer -= i_rewind;
#if defined (MODULE_NAME_IS_i420_rgb)
            CONVERT_YUV_PIXEL(4);  CONVERT_Y_PIXEL(4);
            CONVERT_YUV_PIXEL(4);  CONVERT_Y_PIXEL(4);
            CONVERT_YUV_PIXEL(4);  CONVERT_Y_PIXEL(4);
            CONVERT_YUV_PIXEL(4);  CONVERT_Y_PIXEL(4);
#elif defined (MODULE_NAME_IS_i420_rgb_mmx)
            __asm__( MMX_INIT_32
                     : : "r" (p_y), "r" (p_u), "r" (p_v), "r" (p_buffer) );

            __asm__( ".align 8"
                     MMX_YUV_MUL
                     MMX_YUV_ADD
                     MMX_UNPACK_32
                     : : "r" (p_y), "r" (p_u), "r" (p_v), "r" (p_buffer) );

            p_y += 8;
            p_u += 4;
            p_v += 4;
            p_buffer += 8;
#endif
        }
        SCALE_WIDTH;
        SCALE_HEIGHT( 420, 4 );

        p_y += i_source_margin;
        if( i_y % 2 )
        {
            p_u += i_source_margin_c;
            p_v += i_source_margin_c;
        }
    }
}

/* Following functions are local */

/*****************************************************************************
 * SetOffset: build offset array for conversion functions
 *****************************************************************************
 * This function will build an offset array used in later conversion functions.
 * It will also set horizontal and vertical scaling indicators.
 *****************************************************************************/
static void SetOffset( int i_width, int i_height, int i_pic_width,
                       int i_pic_height, vlc_bool_t *pb_hscale,
                       int *pi_vscale, int *p_offset )
{
    int i_x;                                    /* x position in destination */
    int i_scale_count;                                     /* modulo counter */

    /*
     * Prepare horizontal offset array
     */
    if( i_pic_width - i_width == 0 )
    {
        /* No horizontal scaling: YUV conversion is done directly to picture */
        *pb_hscale = 0;
    }
    else if( i_pic_width - i_width > 0 )
    {
        /* Prepare scaling array for horizontal extension */
        *pb_hscale = 1;
        i_scale_count = i_pic_width;
        for( i_x = i_width; i_x--; )
        {
            while( (i_scale_count -= i_width) > 0 )
            {
                *p_offset++ = 0;
            }
            *p_offset++ = 1;
            i_scale_count += i_pic_width;
        }
    }
    else /* if( i_pic_width - i_width < 0 ) */
    {
        /* Prepare scaling array for horizontal reduction */
        *pb_hscale = 1;
        i_scale_count = i_width;
        for( i_x = i_pic_width; i_x--; )
        {
            *p_offset = 1;
            while( (i_scale_count -= i_pic_width) > 0 )
            {
                *p_offset += 1;
            }
            p_offset++;
            i_scale_count += i_width;
        }
    }

    /*
     * Set vertical scaling indicator
     */
    if( i_pic_height - i_height == 0 )
    {
        *pi_vscale = 0;
    }
    else if( i_pic_height - i_height > 0 )
    {
        *pi_vscale = 1;
    }
    else /* if( i_pic_height - i_height < 0 ) */
    {
        *pi_vscale = -1;
    }
}