Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 29 Jul 2002 09:05:37 -0700 (PDT)
From:      John Polstra <jdp@polstra.com>
To:        stable@freebsd.org
Cc:        gnb@itga.com.au
Subject:   Re: Is use of -O2 still deprecated for buildworld in -stable? 
Message-ID:  <200207291605.g6TG5bK4042448@vashon.polstra.com>
In-Reply-To: <200207290447.OAA22147@lightning.itga.com.au>
References:  <200207290447.OAA22147@lightning.itga.com.au>

next in thread | previous in thread | raw e-mail | index | archive | help
In article <200207290447.OAA22147@lightning.itga.com.au>,
Gregory Bond  <gnb@itga.com.au> wrote:
> >   My understanding is that whether or not -O2 *should* produce exactly
> > equivalent code to -O, there are recurrent reports of people having
> > trouble doing a buildworld with -O2 in -stable or in 4.x release
> > trains, where their problems go away when they revert back to -O. 
> 
> The TCP checksum code seems to be the most common culprit - compiling with -O2 
> seems to create checksum routines that fail in certain corner cases, meaning 
> some hosts or some data patterns cause TCP sessions to freeze.
> 
> [I've not seen it myself, but it has been reported here a couple of times.]

I fixed the checksum code in both -current and -stable a few weeks
ago.

FWIW, a kernel I built with "-O2 -fomit-frame-pointer" for a
network-intensive configuration worked fine.  For my application
(generating and servicing zillions of web requests using special
netgraph nodes I made) these options increased performance by about 9%
compared to the same kernel built with "-O".

John
-- 
  John Polstra
  John D. Polstra & Co., Inc.                        Seattle, Washington USA
  "Disappointment is a good sign of basic intelligence."  -- Chögyam Trungpa


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?200207291605.g6TG5bK4042448>