RE: Linux (fwd)

Adam Haberlach (HaberlaA@testlab.orst.edu)
Wed, 30 Apr 1997 21:07:51 -0700


-----BEGIN PGP SIGNED MESSAGE-----

- -----Original Message-----
From: Justin Dolske [SMTP:dolske@cis.ohio-state.edu]
Sent: Wednesday, April 30, 1997 8:39 PM
To: deschall@gatekeeper.megasoft.com
Subject: Re: Linux (fwd)

On Wed, 30 Apr 1997, Andrew James Alan Welty wrote:

>> Doesn't hurt anything but it will be a little faster just
>running
>> one copy of the client. If there started too close together
>they'll
>> end up with the same keys to check.
>Has anyone actually tried this? I'd be somewhat surprised if it
>was true.

>Every time you context switch, there's a big penalry on the
>processor.
>Pipelines clear, caches flush, and alot of state must be saved.
>Having 4
>processes compeeting for CPU time will only make this happen
>more often. A
>good scheduler will reduce the number of context switches when
>it notices
>that only one program is really compeeting for CPU time.

Next question: Does Linux have a good scheduler? I know that
the SMP scheduler is pretty simple.

ObAnswerToQuestionThat'sBeenAnsweredSixTimes: Only run one
process unless you're running on a multiprocessor system. One
Deschall process will fill the processor because it has no
hardware interaction (except for printing the occasional period)
to cause it to be in a state that does not require the
processor. Two deschall processes will cause
competition--swapping back and forth, and getting slightly less
then 1/2 of a proc each (slightly less is caused by the
switching action itself).

- ---
Adam Haberlach haberlaa@ucs.orst.edu
http://www.engr.orst.edu/~haberlad
Crack DES now! http://www.frii.com/~rcv/deschall.htm

-----BEGIN PGP SIGNATURE-----
Version: 2.6.2

iQCVAwUBM2gWaBSmwqZasiJdAQHMCgP/Z5tVK5NkXSxRAyCmCeO1L30YJga26yKq
t8P5DQHL5U+FR6R8xHBeIJC/1GGH8xqMbz535LMeU89WNMtsQ04Ey3b/MJhRPzve
OofaZe1wNM6NzNcSGSnIc7yHp2vIDLlS66yK5q4XXZDFIgLIqkap3uHoWndJBN1V
X8QdoEyKFTI=
=Wk1f
-----END PGP SIGNATURE-----