Re: [rtl] Re: *Fast* IRQ response ( 2.08us )

[Ken Teh <kmt@slowpoke.phy.anl.gov>]

On Wed, 15 Apr 1998, Daniele Lugli wrote:

> C. W. Wright (1698) wrote:
> 
> > Greetings,
> >
> >    I just measured the time from the leading edge of the IRQ signal to the
> > leading edge of a pulse generated by my rt-linux handler at 2.08us most of
> > the time and sometimes just under 4us.   The measurements made using a
> > Lecroy 9384 1gHz digital scope with one probe on the irq signal and the
> > other on the LPT port I'm pulsing.
> >


Thought I throw in my 2 bits.  I did a similar test last year and measured
an IRQ latency of 1.5us with a jitter of about 1us.  The setup was quite
simple:  an ISA card with an Intel 8255 chip.  The IRQ handler pulses a pin
on the 8255 card in response to an interrupt signal which I can feed into
the card.

I did one more thing which you might be interested in:  I also measured the
task dispatch latency, that is, instead of the IRQ handler pulsing the
output, it dispatches a task which pulses the output bit.  This latency is
about 12us, but the spread is quite large, that is, I see latencies of up to
28us.

The choice of a realtime OS for my project was sealed after I obtained these
results.  I know of no other RTOS that can match these numbers.

Ken Teh

--- [rtl] ---
For more information on Real-Time Linux see:
http://rtlinux.cs.nmt.edu/