Michel D�nzer wrote:
>
> Erwin Rol wrote:
>
> > I am running a G400MAX now and have simmulair problems, but not as bad.
> > Normal latency is around 10usecs, but when a lot of output is writen to
> > a X-terminal it jumps to 500usec. When enabling PCIretries (so writing to
> > the FIFO without checking if it is full) it always shows a latency of
> > 500usec when for example running 'x11perf -all', so it looks like in the MGA
> > driver there is a spot where there is forgotten to check if the FIFO is full
> > (i am looking into the source at the moment).
>
> At least in the glint driver, it's the whole point of PCI retry that the
> driver doesn't even care about the FIFO. Don't know about the other drivers.
Exactly and so is it in the other drivers. I don't know if you followed
the earlier
discussion here, there the cause of the latency was explained like this.
When not checking to see if the FIFO is full and just blindly write in
it, the CPU will stall
when the FIFO is full,untill there is room again. This might be the
fastest way for overall
performance but it causes high ( order of 500 - 5000usec) latency for
real time work.
(atleast that is what is suspected, no other valid explanations where
mentioned )
- Erwin
>
> Michel
>
> --
> Earthling Michel D�nzer (MrCooper) \ CS student and free software enthusiast
> Debian GNU/Linux (powerpc,i386) user \ member of XFree86 and The DRI Project
-- [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/
