> > common case by eliminating a PCI read). A write to a full FIFO will
> > stall the PCI bus (and the rest of the machine) - this condition can
> > last for many milliseconds.
>
> Many milliseconds?
> Do you have an example hardware to test?
You can see how it would take many milliseconds by looking at the amount
of data that must be moved. For example, when dragging an 800x600x32
window, every blit has to move 1.8MB. When I tested this a few years
ago (with early AGP video hardware) this sort of thing could take
25-30ms.
I'd expect the blit bandwidth of a modern video card to be very high,
but of course somebody could easily be dragging a 1300x1000x32 window
nowdays.
Mike Jones and I wrote up some observations about this a few years ago.
This was with a 3D Labs Permedia 2 video board, but we observed the same
problem on several other cards. I don't have access to that hardware
anymore, unfortunately.
http://www.cs.virginia.edu/~jdr8d/papers/hotos7/hotos7.html
John
-- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR
echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED]
---
For more information on Real-Time Linux see:
http://www.rtlinux.org/rtlinux/
