Victor Yodaiken wrote:
>
> On Wed, Jul 07, 1999 at 04:16:44PM +0000, Paolo Mantegazza wrote:
> > Victor Yodaiken wrote:
> >
> > > the disadvantage of one-shot on uniprocessor x86 machines is that
> > > reprogramming the clock takes 10 microseconds or so.
> >
> > That is not true, provided you do not want to use 486s, but is due to
> > the type of implementation of the concept adopted in RTL. In fact I can
> > prove that with a different implementation, e.g our older variant for
> > RTL based on 2.0.xx and RTAI for 2.2.xx, it can go down to 2-3 us on ony
> > PENTIUM and compatible. In that case you can use the CPU clock and have
> > just to do two "outb-s" to reprogram the 8254 timer chip counter for the
>
> You can prove whatever you want, but I have timed a single outb to the isa
> interface at 12microseconds on a PII LX chipset.
Dear Victor,
let have it this way.
Under RTL the you can run, on a PPRO 200, a single bit toggling task at
up to 120 Khz in periodic mode, but in the oneshot mode at 20 Khz it is
short of breathing.
Under our RTL variants, and RTAI, that is roughly 120/95, and the ratio
is the same for all the machines on which I've worked on, from plain
Pentiums upward. Be sure that most of them were of the kind of the
lowest costs, Asian, or Asian like, clones. For instance the dual SMP
SOLTEK motherboards we are using cost less than $180. So no high ranking
brand.
That provided if you cannot buy PCs that are better than mine it means
that it is likely that I have to stop complaining about my humble
budgets.
Please note that figures are just the right order of magnitude, not
meant to be very precise at a few percents.
Ciao Paolo.
--- [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/
