*BSD News Article 54240


Return to BSD News archive

Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!yarrina.connect.com.au!news.uwa.edu.au!classic.iinet.com.au!swing.iinet.net.au!news.uoregon.edu!news.emf.net!overload.lbl.gov!lll-winken.llnl.gov!hookup!news.mathworks.com!tank.news.pipex.net!pipex!sunsite.doc.ic.ac.uk!nntp0.brunel.ac.uk!strath-cs!not-for-mail
From: nbc@muir-10.cs.strath.ac.uk (Neil Clark)
Newsgroups: comp.unix.bsd.freebsd.misc,gnu.gcc.help,comp.os.linux.misc
Subject: Re: gcc optimisations when compiling the kernel
Date: 2 Nov 1995 14:48:39 -0000
Organization: Comp. Sci. Dept., Strathclyde University, Glasgow, Scotland.
Lines: 21
Message-ID: <47alo7$153@muir-10.cs.strath.ac.uk>
References: <478mtj$e2v@plato.ucsalf.ac.uk>
NNTP-Posting-Host: muir-10.cs.strath.ac.uk
Xref: euryale.cc.adfa.oz.au comp.unix.bsd.freebsd.misc:8387 gnu.gcc.help:13746 comp.os.linux.misc:68895

Mark Powell <mark@plato.ucsalf.ac.uk> wrote:
>Looked in the handbook and the FAQ and couldn't find anything in there on
>which optimisations to use when compiling the kernel. 
<SNIP!>
>Comments welcome.

I've tried many variations of the optimization options, but haven't really
come to any solid conclusions. What I can say is that -funroll-loops increases 
the size of the kernel quite significantly, and that doing a simple -O3
doesn't compile. On my NetBSD 68K kernel, I inlined the functions (and kept
them to avoid undefined symbols), as well as unrolling loops. As a result,
the kernel grew from ~500M to ~750M. It *seems* a bit faster, but it could
just be my imagination. Perhaps the same would apply to the FreeBSD
kernel?

At the moment I've settled for the defaults on the trusting presumption
that "they know best." It would be interesting if a battery of kernel
tests were put together to give some indication of performance, although
we all know about the reliability of benchmarks...

Neil