From owner-freebsd-chat Tue May 1 14: 9:10 2001 Delivered-To: freebsd-chat@freebsd.org Received: from odin.ac.hmc.edu (Odin.AC.HMC.Edu [134.173.32.75]) by hub.freebsd.org (Postfix) with ESMTP id 1413537B422 for ; Tue, 1 May 2001 14:09:07 -0700 (PDT) (envelope-from brdavis@odin.ac.hmc.edu) Received: (from brdavis@localhost) by odin.ac.hmc.edu (8.11.0/8.11.0) id f41L94v04088; Tue, 1 May 2001 14:09:04 -0700 Date: Tue, 1 May 2001 14:09:04 -0700 From: Brooks Davis To: Alex Zepeda Cc: chat@freebsd.org Subject: Re: BSD libc for Linux? Message-ID: <20010501140904.A1609@Odin.AC.HMC.Edu> References: <20010430172143.A9910@Odin.AC.HMC.Edu> <002c01c0d25d$10af76a0$08cc1f40@pdq.net> <20010501104642.A27136@Odin.AC.HMC.Edu> <20010501134850.A1175@zippy.mybox.zip> Mime-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-md5; protocol="application/pgp-signature"; boundary="9amGYk9869ThD9tj" Content-Disposition: inline User-Agent: Mutt/1.2.5i In-Reply-To: <20010501134850.A1175@zippy.mybox.zip>; from jazepeda@pacbell.net on Tue, May 01, 2001 at 01:48:50PM -0700 Sender: owner-freebsd-chat@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org --9amGYk9869ThD9tj Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Tue, May 01, 2001 at 01:48:50PM -0700, Alex Zepeda wrote: > On Tue, May 01, 2001 at 10:46:42AM -0700, Brooks Davis wrote: >=20 > > We don't use randomness anywhere in this code so that's not it. >=20 > Well, what *IS* the behavior that's different only on Linux? Significant variations in numerical results. There is some (expected) varition between most platforms, but it is generally of a very low order and mostly shows up in the form on numbers that are nearly zero being bogusly written as negative zero when printed with a fixed decimal. That's expected. What I'm seeing is things like 2.893 vs 2.632 which is a really huge variation. My suspicion is that there's a bug in the libm part of glibc, but there's really no good way to verify that, that I know of. I was hoping that by relacing glibc with something else I could at least determine if that's correct. Eventualy, I will have to go put in a shitload of trace output but that's promices to be exceeding painful given that I've got about 17K lines of application code to instrument and if it's a library problem, I'll probably have to dig into 30K lines of supporting code that provides proof that you can write FORTRAN code in any language. ;-) -- Brooks --=20 Any statement of the form "X is the one, true Y" is FALSE. PGP fingerprint 655D 519C 26A7 82E7 2529 9BF0 5D8E 8BE9 F238 1AD4 --9amGYk9869ThD9tj Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.0.4 (GNU/Linux) Comment: For info see http://www.gnupg.org iD8DBQE67yXwXY6L6fI4GtQRAkkPAKDQFXh4Y/+Aog+uxXzeShUN9YWzGQCfWfEM 5NtRs7jp7GvDqOmR64SWYp0= =Kv1F -----END PGP SIGNATURE----- --9amGYk9869ThD9tj-- To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-chat" in the body of the message