On Sat, Oct 23, 2010 at 1:22 PM, Patrick McNamara <[email protected]> wrote:
> As it currently sits, the framebuffer driver is decent enough with the
> exception of full screen scrolling.  That is very painful.

This is in part because PCI reads take about 4x as long as writes.
The writes get combined into bursts (MTRR setting), while reads are
separate transactions.

A dirty fix for this would be to avoid PCI reads entirely.  It would
be cool if we could somehow get fb to write to the card and to a local
buffer, while doing all reads from the local buffer.  We could do this
by wrapping every fb call or somehow using XAA, EXA, or UXA and just
making all the right fb calls.  Anything that does writes would get
called twice, once for the back buffer, once for the screen.  All
reads would always be directed at the back buffer.

-- 
Timothy Normand Miller
http://www.cse.ohio-state.edu/~millerti
Open Graphics Project
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to