[Kst] What do view objects really need?

George Staikos staikos at kde.org
Wed Jul 21 20:19:14 CEST 2004


Your call... I can't reproduce it anyway so I can leave it where it is now.  
I'll drop the requirement to 8bpp - it can't swap color maps because we blt 
onto the big screen-depth-pixmap in the end.  Just tell me if you see the 
problem again, otherwise I'll leave things where they are.

On Wednesday 21 July 2004 13:16, Barth Netterfield wrote:
> On July 19, 2004 10:50 pm, George Staikos wrote:
> > As far as the pixmap goes, what do view objects really need?  I just
> > committed a change that makes them 16bpp, and I was even considering
> > going as low as 8bpp.  Does this make sense?  Are we even making use of
> > this much colour space in each individual view object?  I think we can
> > survive with 8bpp in most cases, especially once labels are turned into
> > proper view objects.  Then I would consider making the depth of the
> > buffer a parameter to the backbuffer class (labels will want something
> > more than 8bpp to have cleaner antialiasing I think).
>
> I'm guessing 8bpp is OK as long as we don't end up with color map swapping
> when we enter and leave the window (ancient effect on old low memory 8bpp X
> displays), esp if it speeds things up.
>
> > Also, Barth, does the MemoryOptim make a difference for you on the
> > machine that experiences pixmap pressure?
>
> Don't know.  It may have been an ATI bug; since I moved from ATI's binary
> to the OSS driver, all has been well.  I can try again with the ATI driver.
>
> > Trying to reduce the pixmaps was the
> > best that Jim Gettys could come up with when we talked about it for a few
> > minutes. He doesn't think hoping for client side rendering to help is
> > very realistic (or ideal).  If MemoryOptim doesn't help, I'd like to
> > convert it to something more buffered.
>
> Or we can warn against ATI's driver and get on with our lives (?).

-- 
George Staikos
KDE Developer				http://www.kde.org/
Staikos Computing Services Inc.		http://www.staikos.net/



More information about the Kst mailing list