On Thu, Nov 18, 1999 at 02:43:06AM +0100, Bernhard Kuhn wrote:
> > Otherwise, are you sure you are recompiling 3dfx-module?
>
> yes, absolutly!
>
Oh, well.
> > This is probably a simple error. We forbid cross CPU changes of period, but
> > you report what may be the result of the cpu field on the task not being
> > set properly.
>
> So, i have to bind a task to a specific CPU to be able to change
> the period within the task?
The way it should work is that the cpu is set to a default cpu if you
don't specify it. I'm speculating that this is not being done right
because I don't want to look at the code right now.
> IMHO, at least it it a good idea to not let a rt-task jump
> across several CPU, since the not-well deterministic
> intel-architecture would be even less deterministic
> due to cache-flushes and MESI-Protocol.
Absolutely. We do not do load balancing for the same reason we don't try
to calculate priorities at run time.
> Although, from technical point of view, it would be more
> interessting to be able to fix the task to a subset of CPUs,
> like it is in RTAI/RTHAL :-)
Open source allows one to make these choices.
> [reiserfs]
> >
> > That is cool, but it will encourage poor programming practice by
> > kernel programmers since the consequences are less -)
>
> Are you talking about the guys in redmond? :-)
I was talking about myself.
> BTW: if it would help, i can give you remote-access to my
> system to determine the problem with the 3dfx-module.
> IMHO, the "setperiod across CPUs"-Problem is only due to
> wrong programming ... i will check that later.
I'll email later and see if I can dig something up here.
--- [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/
