[cairo] glitz - YUV - _glitz_pixel_transform_op_t stride

[David Reveman]

Hi Patrick,

On Mon, 2006-02-06 at 19:44 +0000, Patrick Pailhes wrote:
> Hi there,
> 
> I have looked at the code in glitz_pixel.c about YV12/RGB conversion
> (notably the functions _fetch_yv12, _store_yv12 and
> _glitz_pixel_transform). I want to implement and test the functions
> _fetch_imc2 and _store_imc2 that would perform respectively IMC2/RGB
> and RGB/IMC2 conversions. The IMC2 format is very similar to YV12. The
> difference is that the V and U lines are interleaved at half stride
> boundaries.
> 
> Now, in terms of code, the structure glitz_pixel_transform_op_t holds
> the Y, U and V components (see below):
> 
> typedef struct _glitz_pixel_transform_op {
>     char                 *line, *line2, *line3;
>     int                  offset;
>     glitz_pixel_format_t *format;
>     glitz_pixel_color_t  *color;
> } glitz_pixel_transform_op_t;
> 
> For YV12, Y is stored in line, V in line2 and U in line3. However, 
> for the IMC2 format, both V and U would be stored in the variable
> line2. Now, I believe that we need an extra a variable in the
> structure glitz_pixel_transform_op_t. This variable would hold the
> stride of the YUV array (i.e. the width of the surface). The stride is
> needed in order to retrieve the U components (see below):
> 
> int16_t u = ((uint8_t *)op->line2)[(op->offset >> 1) + (op->stride >>
> 1)] - 128; // u = op->line2[offset/2 + half stride])
> 
> Thus, I think we could add a minor change to the structure
> glitz_pixel_transform_op_t:
> 
> typedef struct _glitz_pixel_transform_op {
>     char                 *line, *line2, *line3;
>     int                  offset;
>     int stride;
>     glitz_pixel_format_t *format;
>     glitz_pixel_color_t  *color;
> } glitz_pixel_transform_op_t;

This is OK.

You can use the same fragment program for IMC2 as for YV12 surfaces,
right? Only difference is that the IMC2 data will loaded to a texture
using two glTexSubImage2D calls instead of three.


-David