Date: Fri, 10 Nov 95 09:53:31 -0800 From: Bakul Shah <bakul@netcom.com> To: "Ron G. Minnich" <rminnich@sarnoff.com> Cc: Larry McVoy <lm@slovax.engr.sgi.com>, hackers@freebsd.org, "William A. Arbaugh" <waa@aurora.cis.upenn.edu>, Theo de Raadt <deraadt@theos.com>, Chuck Cranor <chuck@maria.wustl.edu> Subject: Re: larry: you might want to add this to lmbench (but i'm not sure) Message-ID: <199511101753.JAA10892@netcom22.netcom.com> In-Reply-To: Your message of "Fri, 10 Nov 95 11:15:34 EST." <Pine.SOL.3.91.951110110255.13530A-100000@morse>
next in thread | previous in thread | raw e-mail | index | archive | help
What count did you use to get these numbers?
Running it on a SUN 3/50 with SunOS3.5 yields about 1
ms/write(); not bad considering the syscall overhead!
On a 25Mhz 486 running NetBSD-1.1A I get about 17ms/call
(wall clock). But system time is only about 1.3ms/call.
Also, there was a lot of disk activity.
On a 100MHZ P5 running FreeBSD-2.1 I get 0.287ms/call and
system time was also about the same.
> but i'm not sure why freebsd has to be so slow. I'm willing to blame it
> on the mach vm, since it has been such a problem in so many other ways.
For the {Free,Net}BSD case you seem to be measuring
something totally different from the rest that do arg
checking upfront. In the latter case you are measuring
time to reject an invalid write() syscall.
I don't see what system parameter your benchmark measures
(other than pointing out if arg checking is done early or
late). Since all syscall args need to be checked sooner or
later, they may as well get checked upfront.
> For some work i'm doing it is measuring an important value however.
What parameter are you trying to measure? Hmm... perhaps
you are mapping in pages when you discover an invalid
address?
--bakul
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199511101753.JAA10892>