Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 24 Jul 2000 13:29:11 -0400 (EDT)
From:      mi@aldan.algebra.com
To:        John Baldwin <jhb@pike.osd.bsdi.com>
Cc:        stable@freebsd.org
Subject:   Re: Recommended compilation optimizations
Message-ID:  <200007241729.NAA34094@misha.privatelabs.com>
In-Reply-To: <200007241707.KAA49181@pike.osd.bsdi.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On 24 Jul, John Baldwin wrote:
= > = -O -pipe
= > =
= > = AFAIK that's the only optimization that is more or less guaranteed
= > = to work.
= > 
= > I've  seen  '-fexpensive-optimizations'  break havoc  in  the  squid
= > binary,  so  I  no  longer  use it  for  anything,  but  '-mcpu=i686
= > -march=i686' seems fine to me and  tells the compiler that your only
= > target processor  is of the  686 class. If  you are not  planning to
= > debug the thing, add '-fomit-frame-pointer'.
= > 
= > 	-mi (who gets  annoyed every  once in a  while when  he  sees
= > 	     people discouraging optimization as dangerous instead of
= > 	     fixing the compiler)
= 
= Uh,  have you  actually  _looked_ at  the gcc  source?  Making such  a
= statement is  rather arrogant if you  aren't willing to put  some work
= into the compiler yourself.  gcc is not an easy program  to fix, it is
= rather large and complex.

I was  not blaming you, neccessarily.  As for me,  I did file a  few PRs
with FreeBSD  and/or with GCC  people, providing the exact  C-code, that
breaks and more details. I think, it is important for everyone bitten by
some brokennes of some aggressive optimization in GCC to investigate and
file the bug reports... This is actually, true about any brokenness :)

In the above  mentioned case of squid, the reduced  C-code snippet, that
continued to brake on FreeBSD was  compiled properly on Mandrake (by the
same version of EGCS), so it looks like a FreeBSD-specific problem.

Instead of encouraging bug-reports, the attitude has so far been: "well,
if you use high optimization -- you  are on your own", as exemplified in
the follow-up to my i386/19245 . As a result, at least, in this example,
the same thing works on Linux, that does not on FreeBSD.

	-mi




To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-stable" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200007241729.NAA34094>