"Bruce Anderson" <brucea@spacestar.net> writes:
> See: <http://www.ece.cmu.edu/~koopman/pc_cache/espcache.html>
>
> "Perils of the PC Cache"
>
> Philip J. Koopman, Jr 
>
> Draft text for article published in: 
> Embedded Systems Programming, 6(5), May 1993, pp. 26-34 
>
> also: "Embedded Computer Systems"
> <http://www.ece.cmu.edu/~koopman/embedded.html>

Sadly, none of that answers his question at all.

As for the actual question that was asked: in practice, NetBSD
generally does fine for "soft" realtimish things like doing video
playback, but it doesn't have the sort of actual realtime scheduler
extensions one wants, and there are drivers and other portions of the
kernel that sometimes lock up a machine for unacceptably long periods
of time in busy waits. I'm not sure NetBSD is deployable in tasks that
require tens of microsecond of latency without careful engineering on
the part of the application designers.

Perry


> On Sat, Nov 29, 2003 5:37 AM, Elder Costa <mailto:elder.costa@terra.com.br>
> wrote:
>>Hi. After googling, browsing through documentation and mailing list 
>>archives, I couldn't find anything about the subject.
>>
>>I would like to know how NetBSD performs as far as determinism is 
>>concerned. That is, are there mesurements on task latency from interrupt 
>>events? Is NetBSD deployable for jobs that require latency of 
>>microsseconds/tens of microsseconds at most?
>>
>>Pointers to articles, faqs, docs etc. are welcome and will be very much 
>>appreciated.


-- 
Perry E. Metzger		perry@piermont.com