Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 15 Apr 1998 09:39:47 -0400
From:      sbabkin@dcn.att.com
To:        tarkhil@asteroid.svib.ru, dyson@FreeBSD.ORG
Cc:        hackers@FreeBSD.ORG, smp@FreeBSD.ORG
Subject:   RE: SMP performance decrease?
Message-ID:  <C50B6FBA632FD111AF0F0000C0AD71EEACDEB1@dcn71.dcn.att.com>

next in thread | raw e-mail | index | archive | help
> ----------
> From: 	John S. Dyson[SMTP:dyson@FreeBSD.ORG]
> 
> Alexander B. Povolotsky said:
> > Hello!
> > 
> > I've just benchmarked (by bytebench from ports) my new server, with
> 64M RAM
> > (for now; will increase soon) and 2 PII-233 processors (EliteGroup
> p6LX2-A
> > mainboard).
> > 
> > Here are the results:
> 
May be some description of tests can help:

> > 
> > Test                            Non-SMP      SMP
> > arith (double)                   27.5        28.1 
> 
This test is running in one process so it hardly can show anything.
An interesting modification would be to run 2 parallel processes.

> > Drystone 2 (no reg. vars)        23.6        24.1
> 
The same. The speed gain may be because other processes that are
running in background are using the second processor now.

> > Execl thoughput                  90.8        46.8   <- ??? 
> 
I can't remember about this

> > File copy                        35.1        37.7   
> 
Also single-threaded.

> > Pipe-based context switching     20.4         9.0   <- ???
> 
Two processes and a pipe, passing 1 byte each time in alternating
direction. The number of bytes passes is supposed to be the same
as the number of context switches. In fact, the processes here
are running _not_concurrently_, they are synchronized by byte passing.
So the speed can hardly be higher, but should not be twice
slower either until there is high interprocessor synchronization
overhead compared to context switch overhead (provided that FreeBSD 
context switch overhead is really low, that may be the case).

> > Shell scripts (8 concurrent)     12.8        12.2   <- ???
> 
8 shell scripts are running concurrently. This is supposed to
simulate 8 users running. And think that this is a somewhat
weird result.

Serge



To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?C50B6FBA632FD111AF0F0000C0AD71EEACDEB1>