Date: Tue, 29 Oct 2002 13:06:36 -0800 (PST) From: Matthew Dillon <dillon@apollo.backplane.com> To: Terry Lambert <tlambert2@mindspring.com> Cc: Raymond Kohler <raymond.j.kohler@lmco.com>, current@FreeBSD.ORG Subject: Re: questions about the state of current Message-ID: <200210292106.g9TL6aoc010659@apollo.backplane.com> References: <2570443.1035916854787.JavaMail.wshttp@emss03g01.ems.lmco.com> <3DBEF55E.A0F9ED1B@mindspring.com>
next in thread | previous in thread | raw e-mail | index | archive | help
:Most of the speed difference is WITNESS, INVARIANTS, and other
:debugging code that's turned on by default in the config files
:for -current. You can turn most of it off. That said, -current
:is slower than -stable in a number of places, so expect some
:slowdown, if you are running non-concurrent code.
I would concur with this diagnosis. With witness turned off
-current is around 15% slower then -stable for general purpose
computing, like a 'make buildworld -j 20', and I expect that
-stable will beat out -current on single-cpu boxes for a long
time to come. That said, it should be noted that nearly all the
really cool development projects are only happening in -current.
Things like KSEs, hardware crypto support, UFS2, and so forth,
only exist in -current. Very little of this work is going to be
MFC'd so depending on your needs -current could very well end up
faster.
And, of course, there is the fact that computing power seems to
double every year. Since -current's overhead is in large part
due to mutexes and other concurrency mechanisms, and these are
literally pure-cpu mechanisms rather then memory or I/O dependant,
decisions should be based on capability rather then something as
insignificant as a 15% performance difference between the 'rough cut'
-current and the well aged -stable. In coming years concurrency
is going to become the leading performance-improving mechanism
for computers.
-Matt
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200210292106.g9TL6aoc010659>