Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 10 Oct 2002 22:20:01 -0700
From:      Steve Kargl <sgk@troutmask.apl.washington.edu>
To:        Don Lewis <dl-freebsd@catspoiler.org>
Cc:        gallatin@cs.duke.edu, kabaev@bellatlantic.net, mb@imp.ch, current@FreeBSD.ORG
Subject:   Re: HEADSUP: GCC 3.2.1 update is coming
Message-ID:  <20021011052001.GA73095@troutmask.apl.washington.edu>
In-Reply-To: <200210110336.g9B3aJvU043765@gw.catspoiler.org>
References:  <20021010232253.GA68378@troutmask.apl.washington.edu> <200210110336.g9B3aJvU043765@gw.catspoiler.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Oct 10, 2002 at 08:36:19PM -0700, Don Lewis wrote:
> On 10 Oct, Steve Kargl wrote:
> > On Thu, Oct 10, 2002 at 07:10:44PM -0400, Andrew Gallatin wrote:
> >> 
> >> This is a terrible example in that its impossible to tell what
> >> happened, but it sure shows that something is wrong.
> >> 
> >> Is there a floating point regression suite that you can point me at?
> >> That would be more useful for debugging.  
> >> 
> > 
> > ftp://ftp.netlib.org/fp/ucbtest.tgz
> > 
> > This is Kahan and his students floating point testsuite.
> > It normally check for conformance to IEEE-754.
> 
> For something a bit older, see <http://www.netlib.org/paranoia/>.
> 
> 

Paranoia only checks things like rounding and epsilon.
ucbtest tests corner cases for the standard intrinsics.
ucbtest should be a much better test of libm than
paranoia.

-- 
Steve

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?20021011052001.GA73095>