From mboxrd@z Thu Jan  1 00:00:00 1970
Received: (from weis@localhost) by pauillac.inria.fr (8.7.6/8.7.3) id NAA28574 for caml-redistribution; Fri, 26 Feb 1999 13:15:38 +0100 (MET)
Received: from nez-perce.inria.fr (nez-perce.inria.fr [192.93.2.78]) by pauillac.inria.fr (8.7.6/8.7.3) with ESMTP id RAA06011 for <caml-list@pauillac.inria.fr>; Thu, 25 Feb 1999 17:57:37 +0100 (MET)
Received: from tequila.cs.yale.edu (TEQUILA.CS.YALE.EDU [128.36.229.152])
	by nez-perce.inria.fr (8.8.7/8.8.7) with ESMTP id RAA13976
	for <caml-list@inria.fr>; Thu, 25 Feb 1999 17:57:35 +0100 (MET)
Received: from tequila.cs.yale.edu (localhost [127.0.0.1])
	by tequila.cs.yale.edu (8.8.7/8.8.7) with SMTP id LAA22179
	for <caml-list@inria.fr>; Thu, 25 Feb 1999 11:57:35 -0500
To: caml-list@inria.fr
Sender: weis
From: Stefan Monnier <monnier+lists/caml/news/@tequila.cs.yale.edu>
Newsgroups: lists.caml
Subject: Re: bug in floating point implementation ?
References: <199902250043.BAA21786@nez-perce.inria.fr> <19990225131027.37364@pauillac.inria.fr>
Date: 25 Feb 1999 11:57:28 -0500
Message-ID: <5ln222a0av.fsf@tequila.cs.yale.edu>
X-Newsreader: Gnus v5.5/Emacs 20.3
Path: tequila.cs.yale.edu
NNTP-Posting-Host: tequila.cs.yale.edu
X-Trace: 25 Feb 1999 11:57:28 -0500, tequila.cs.yale.edu

>>>>> "Xavier" == Xavier Leroy <Xavier.Leroy@inria.fr> writes:
> The Intel x86 has an interesting quirk: it computes internally in
> extended precision (80 bits), and rounds to double precision (64 bits)
> only when storing a result in memory.  This means that in a piece of

But you can make it work in 64bit-mode (or 32bit-mode for that matter)
at which point the discrepencies between true-real 64bit and x86-64bit become
very small (mostly due to the exponent still allowing a wider range).
Of course, there might be good reasons for not using this mode-switch.


	Stefan