Hello,
I am currently measuring the interrupt latency on a Pentium (586)
running at 200 MHz running Linux 2.0.36 with the real time Linux
extensions installed. The scheme is to trigger an interrupt by feeding a
pulse to pin 10 on the parallel port and trigger a digital scope with this
pulse. Channel 2 of the scope monitors bit 0 (pin 2) of the parallel port.
The only response to the interrupt is to set bit 0 high for about 100
microsec and then set it low again and hybernate until the next interrupt
comes along. The scope has a so-called envelope mode which allows the
extremes of the waveforms to be easily seen. As I write this message the
latest test has been under way for 10 hours.
On an unoccupied system the total latency varies from 2 to 6
microsec. However I am interested in the upper limit so I exercise the
regular Linux OS strenously during the tests. Currently I transmit TCP/IP
packets over a 100 base T line to another dedicated system which happens
to run Win NT. The typical thruput on the Network is 9 MB/s. (This rate is
barely affected by the presence or absence of the interrupts.) The maximum
latency I have seen while the network transfers are going on is 22
microsec.
That's the good news. The bad news is that the system has a
definite tendency to crash after a few hours of operation. It is much too
early to blame this on any particular thing. I suspect something to do
with memory swapping. The system has 64MB of RAM and 127 MB of swap space.
The "top" utility shows only 3 MB of RAM free and only 4 K of the swap
space used. Yesterday I was able to trigger a crash just by playing around
in the control panel (or was it a coincidence?).
I plan to set up a web page describing our foray into the real
time Linux world in the very near future. All of our test results will be
posted there (on fermi.jhuapl.edu).
Cheers, bruce
---
Bruce L. Gotwols
Johns Hopkins University, Applied Physics Lab., Laurel MD 20723
Internet: [EMAIL PROTECTED]
Phone: 240-228-4543 FAX: 240-228-5548
Space Oceanography Group Home Page -- http://fermi.jhuapl.edu
On Wed, 24 Feb 1999, Robert Wilhelm wrote:
> Has anybody measured interrupt latencies on different OS?
> I need some figures to convince my chief to choose
> the right OS...
>
> --
> Robert Wilhelm
> --- [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/
>
--- [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/
