Date: Wed, 14 Dec 2005 19:00:42 -0800 From: Gary Kline <kline@tao.thought.org> To: Chuck Swiger <cswiger@mac.com> Cc: Gary Kline <kline@tao.thought.org>, freebsd-questions@freebsd.org, Kent Stewart <kstewart@owt.com> Subject: Re: fquestions Message-ID: <20051215030042.GA45655@thought.org> In-Reply-To: <43A0C8ED.4090209@mac.com> References: <20051214171014.GB37495@thought.org> <7088318B-3141-44E6-9F50-CB51F6CAE501@mac.com> <20051214211749.GJ41870@thought.org> <200512141342.22051.kstewart@owt.com> <20051215005519.GA44946@thought.org> <43A0C8ED.4090209@mac.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Dec 14, 2005 at 08:37:49PM -0500, Chuck Swiger wrote: > Gary Kline wrote: > > On Wed, Dec 14, 2005 at 01:42:21PM -0800, Kent Stewart wrote: > [ ... ] > > Does it make any sense to use O3 when compiling stuff, > > when stuff includes world/kernel/drivers? Does upping the > > optimization make any significant difference in system > > performance, in other words? Kent? Anybody? > > No. You are likely to vastly increase the amount of time it takes to compile > the system without gaining any performance that's noticable. The system > generally shouldn't be spending a lot of CPU in the kernel, anyway, compared > with the amount of time running user-mode code. (Firewalls and routers are a > significant exception, however.) Good to know, thanks. I have done a lot of tuning on my DNS server. When I upgrade, you've given me more to think about. > > If you want your system to perform better, benchmark the work it's actually > doing, and then tune from there. Spending lots of time to optimize a part of > the system that is already pretty efficient isn't going to do much, whereas > solving the bottleneck will make a useful difference. > For some reason, using Gnome sometimes brings my test server to a crawl. It's running Ubuntu and they say straight out that "Linux is just a kernel." Other than the kernel my test system has the same software as I've got here. Here, I run ctwm and everything's snappy. Are there any benchmark suite you'd recommend to see what's sucking up the most cycles? On every platform, not just my test box (old e-machines:). top is not very helpful. gary -- Gary Kline kline@thought.org www.thought.org Public service Unix
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20051215030042.GA45655>