On Apr 12, 2007, at 11:37 PM, prdcomp wrote:
> Hi,
>
> I've been building (again, after some time) lfs recently. Apart
> from a (solved) glibc testsuite doubt, everything seens to have
> gone well.
>
> So now I've searching about optmization options - including old lfs-
> support mail archive.
>
> I know it can lead to failures - I think the book itself mentions
> it, but are there safe(r) flags worth using?
>
> Recently, I did an experiment: I build lfs (using jhalfs) without
> optmizations, and in a spare partition built it subsequently with
> (mainly) the following flags: "-O2 -pipe -march=athlon-xp". Those
> are rather common options, but the number of unexpected (and quite
> expected by now) libmudflap failures raised from the traditional 6
> to 320!
>
> Actually, I think the real question is: is it worth trying to
> optimize a LFS (initial) system? For BLFS, I tend to believe it is.
> For LFS itself, not that sure.
>
> Any tips/ideas on that folks?
>
> Thank you!
>
> Paulo
Hello Paulo,
Have you bothered to read the gcc online documentation for the gcc
release you are using? Check at gcc/Optimize-Options.html#Optimize-
Options of the gcc release you are using. Just search gcc online
documentation. Also, later into the manual you'll find specific
options for your cpu. The tests aren't ranked for architecture/
machine, but by architecture. You can probably figure out the rest
yourself.
I will answer your question. LFS.. BLFS.. wherever you start your
optimization, whatever depends on that later, will use those
optimizations. Why you tend to believe to start with optimizations
with BLFS is questionable. Things are linked dynamically, mostly, in
LFS and BLFS. Optimizations that are introduced to libraries will be
used in the future with binaries and other libs that link to them.
LFS... just another linux OS build... has nothing to do with
optimizations. I've done a few builds with optimizations. I even
attempted different builds with gentoo safe and experimental
optimizations. -O3 is death. -O2 is aright. -O is fine. I suggest
defaults as the authors intended, and I don't suggest changing
otherwise. However experimentation is most of the fun. Research and
learn from it. Some projects will like optimizations better than
other, but for tests, they aren't going to be biased toward
optimizations. You're issue is that if you use optimizations... will
it break a build because the test suite results are abnormal? Maybe...
Now I stated earlier that tests aren't ranked for architecture/
machine... but I figure some packages have tests for machine, too...
Depends what is searched, detected, and used.
In most instances, it isn't worth it, later on down the line a
headache may be induced, but I leave that to you to find out.
Experiment and find out. I hope you have the time and energy to do
that, it's worth the outcome in experience. There are many articles
that discuss optimizations. I hope you find something that helps you.
Sincerely,
William
--
http://linuxfromscratch.org/mailman/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/lfs/faq.html
Unsubscribe: See the above information page
