Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 26 Mar 2003 22:09:34 +0100
From:      Alexander Leidinger <Alexander@Leidinger.net>
To:        gcc-gnats@gcc.gnu.org
Cc:        ljrittle@gcc.gnu.org
Subject:   Re: optimization/10189: pentium4 breaks suns libm code for __ieee754_pow(double x, double y)
Message-ID:  <20030326220934.398c7455.Alexander@Leidinger.net>
In-Reply-To: <20030326130118.8374.qmail@sources.redhat.com>
References:  <20030326130118.8374.qmail@sources.redhat.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On 26 Mar 2003 13:01:18 -0000
ljrittle@gcc.gnu.org wrote:

> Synopsis: pentium4 breaks suns libm code for __ieee754_pow(double x, double y)

[...]

>  FreeBSD src tree; and (c) that really cares about building FreeBSD
>  src with special CPU settings (do you guys really see enough speedup
>  to warrant this extra nightmare? ;-)

Without knowing anything about the FreeBSD related PRs in the gcc PR
database I just comment on the last part of the quoted sentence...

The official "allowed" optimization is "-O". But it is as easy as
setting 'CFLAGS=-my-special-optim' in /etc/make.conf and start "make
buildworld" in /usr/src to rebuild the userland with new optimizations.
And trust me, as long as gcc ships with a description of other
optimizations beneath "-O" there will be (clueless or smart... does it
really matter here?) people which will try those optimizations on
everything (after I managed to convince the Linux version of icc to
generate FreeBSD object files and committed a port into our ports
collection one of the first questions was "Are we are able to build the
userland/kernel with it?", and now after icc is also able to link files
without the help of gcc they ask "How much does it gain us to build the
userland/kernel with icc?", even if it isn't possible to use icc to
build the entire (or even large parts of the) kernel/userland yet). So
it isn't a matter of "does it improve things if I do it this way" or "is
it possible to do it this way", it's a matter of "how many PRs does it
generate when -march=pentium4 breaks something but other -march=pentiumX
optimizations don't"...

Thanks for your insightful mail,
Alexander.

-- 
            0 and 1. Now what could be so hard about that?

http://www.Leidinger.net                       Alexander @ Leidinger.net
  GPG fingerprint = C518 BC70 E67F 143F BE91  3365 79E2 9C60 B006 3FE7



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030326220934.398c7455.Alexander>