Matthias Buelow wrote:

> Sung N. Cho writes:
>
> >I read this article from LinuxToday,
> >http://linuxtoday.com/news_story.php3?ltsn=2002-03-08-010-20-PS, and
> >recompiled my NetBSD kernel with maximal optimization allowed by GCC and
>
> I'm sure you have also read the comments at the bottom of the
> article and stumbled over this (imho) very sensible reader:
>
> http://linuxtoday.com/news_story.php3?ltsn=2002-03-08-010-20-PS-0008
>
> --mkb

Hi,

I have been using programs (GNOME, KDE, XFCE... all application I use),
including the netbsd system itself compiled with -O3 option and never had
any stability problems so far.  Infact, I had more problem with my netbsd
last week when I compiled the whole system with default optimization options
and had lots of problems!  Maybe the incident had nothing to do with the
default options and I know it isn't.  But, my point is that  I eventually
reinstalled everything from scratch last Thursday with -O3 and it works fine
again.  Last night, I recompiled the whole thing, including the netbsd
system, using -O9 -march=i686 and everything works fine without a sigle
instability.  I'm now beginning to believe that GCC optimizations are
reliable.  Note that I'm not adding anything to -O9.  I think much of the
stability issue comes from the fact that people add more options along with
-O? and accidently add things that shouldn't be added, e.g., with -O option,
you should never add -ffast-math etc.

P.S.  the last time I had building problem was due to the fact that I had
noatime option in fstab.  I wouldn't recommend adding noatime to fstab for
better disk I/O.  Some functions depend on it as I found out when building
the system from source!

Sung