Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 23 Mar 2003 20:22:33 +0100
From:      Till Riedel <till@f111.hadiko.de>
To:        Michael Nottebrock <michaelnottebrock@gmx.net>
Cc:        freebsd-current@freebsd.org
Subject:   Re: libm problem
Message-ID:  <20030323192233.GB2178@f111.hadiko.de>
In-Reply-To: <200303231939.10040.michaelnottebrock@gmx.net>
References:  <20030323182405.GA2135@f111.hadiko.de> <200303231939.10040.michaelnottebrock@gmx.net>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Mar 23, 2003 at 07:39:06PM +0100, Michael Nottebrock wrote:
Content-Description: signed data
> Actually, when I said "better", I really meant "doesn't break -O(ptimized) 
> code" rather than "faster". :)
ok, got you now. it is neither faster nor better.
I tested it with e_pow.c.
funnily though using the pentiumpro proccessor cost table "fixes" the libm
problem :-).
the point still is though, that enabling other things (or like me changing the cost table) doesn' fix anything,
because eg if someone changes some stuff not directly related
to sse2 code selection again may find another rule for sse2 that is
buggy again. 
I think no one should simply use the sse2 stuff at the moment,
because you will notice problems too late. My
mp3s sounded horrible and gv didn't display pdf files correctly.
if there hadn't been libvorbis though that had a division by zero,however, i probably never would have located the bug.

till

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




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030323192233.GB2178>