[H-GEN] SETI at home

[Gearon, Paul]

(Note reply-to: being general at humbug.org.au vs "Gearon, Paul" <pgea at qcom.com.au>)

> >It uses idle CPU cycles and doesn't effect the performance of the
> machine
> at
> >all. It only connects to the internet to upload the results and
> download
> another
> >workunit.
> 
> Correct me if I am wrong, but doesn't seti use up inordinate amounts
> of ram
> when compared with rc5? (another distributed computing client)  Also,
> doesn't seti take *much* longer to complete a block than rc5?
> 
I don't know about the RAM usage, but comparing completion of RC5 blocks
to Seti blocks is completely meaningless.  I'm sure that analysing much
larger RC5 blocks would take longer than analysing a Seti block.  It's
like comparing apples with the price of fish in china  :-)
('scuse the mixed metaphore, but it should illustrate my point).

The comment about idle CPU cycles is a little misleading however.  That
is possible under OSs like Windoze and QNX which only allocate
timeslices to lower priority processes when all the higher priority
processes are blocked (For this reason it is rare that processes change
from the default priority level).  Under Linux, _all_ processes are
allocated timeslices, no matter how much they are niced.  The difference
is that low priority processes only get very small timeslices in
relation to other non-blocked processes.

There was some discussion about this on the kernel-smp list some months
back.  Several people wanted ALL their CPU usage during the day for
scientific tasks, and for RC5 to basically "take over" at night.  A few
people tried to come up with patches, but Linus explained some of the
realities of the Linux kernel and the discussion died off.

Paul Gearon
Catapultam habeo. Nisi pecuniam omnem mihi dabis, ad caput tuum saxum
immane mittam.  (Translation from latin: "I have a catapult. Give me all
the money, or I will fling an enormous rock at your head.") 



--
This is list (humbug) general handled by majordomo at lists.humbug.org.au .
Postings only from subscribed addresses of lists general or general-post.