On Fri, 29 Mar 2002, Keith Packard wrote:

>
> Around 15 o'clock on Mar 29, Vladimir Dergachev wrote:
>
> > I was under impression that Radeon driver only uses 565 and 32bpp ARGB
> > framebuffer formats. I do not think it makes sense to support Xv with 8bpp
> > modes so I thought the code to be correct.
>
> For TrueColor, the radeon driver also supports 555, which is the format I
> was using.
>
> I don't understand why you think 8bpp isn't a reasonable format; many apps
> run only on PseudoColor displays.  The video isn't restricted to that
> format as it bypasses the LUT, so this format would have no effect on the
> appearance of the Xv images.

The reason is that we use colorkeying for clipping. 8bbp display has a
very large chance that this will break down and part of video will show
thru. Additionally, 8bpp is just plain bad for color intensive apps - you
will be ok if your application _only_ displays video via overlay, but what
if it wants to display a snapshot or user interface as well ?

Which apps require 8bpp ? I have not seen one in a long while.. Though I
do stick to the ones with source code available.

>
> > Are you trying to allow pixel values to be passed in the same format as
> > the visual associated to the window you apply XvPutXXXX call to ? If so, I
> > do not understand how your code works..
>
> To work properly, the color key must be programmed with the RGB values it
> will see when the given pixel value is sent through the LUT.  That's why
> the code I wrote uses the current installed colormap, rather than the
> windows colormap.  It occurs to me now that the code must also track
> colormap changes and reload the color key as appropriate.

And this is exactly the reason I did not want to support 8bpp for video
applications in the first place. It is a large hassle and the effort is
much better spent on making it work right with 16bpp and 24bpp.

8bpp and video just don't mix well, be it using XShmPutImage or
XvShmPutImage.

                          Vladimir Dergachev

>
> Keith Packard        XFree86 Core Team        Compaq Cambridge Research Lab
>
>

_______________________________________________
Xpert mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/xpert

Reply via email to