From mboxrd@z Thu Jan  1 00:00:00 1970
X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me
X-Spam-Level: 
X-Spam-Status: No, score=-1.3 required=5.0 tests=BAYES_00,INVALID_MSGID
	autolearn=no autolearn_force=no version=3.4.4
X-Google-Language: ENGLISH,ASCII-7-bit
X-Google-Thread: 107079,c7637cfdf68e766
X-Google-Attributes: gid107079,public
X-Google-Thread: 109d8a,c7637cfdf68e766
X-Google-Attributes: gid109d8a,public
X-Google-Thread: f8362,c7637cfdf68e766
X-Google-Attributes: gidf8362,public
X-Google-Thread: f43e6,c7637cfdf68e766
X-Google-Attributes: gidf43e6,public
X-Google-Thread: 103376,c7637cfdf68e766
X-Google-Attributes: gid103376,public
From: eggert@twinsun.com (Paul Eggert)
Subject: Re: floating point comparison
Date: 1997/08/14
Message-ID: <5t08g3$4jn$1@shade.twinsun.com>#1/1
X-Deja-AN: 268335403
Distribution: inet
References: <mheaney-ya023680002907972345420001@news.ni.net>
 <dewar.870304869@merv> <mheaney-ya023680003007972240570001@news.ni.net>
 <01bc9dca$762b6000$e1e989cc@gort> <dewar.870386556@merv>
 <5s5erh$m92$1@shade.twinsun.com> <dewar.870870402@merv>
Organization: Twin Sun Inc, El Segundo, CA, USA
Newsgroups: 
 comp.lang.ada,sci.math.num-analysis,comp.software-eng,comp.theory,sci.math
Date: 1997-08-14T00:00:00+00:00
List-Id: <comp.lang.ada>


dewar@merv.cs.nyu.edu (Robert Dewar) writes:

> your comment on efficiency on the x86 is technically wrong. You can place
> the x86 floating-point into 64 bit mode, so that there is no efficiency
> penalty, and indeed at least some C compilers on the x86 do this.

But in 64-bit mode, the x86 doesn't round denormalized numbers properly.
It simply rounds the mantissa at 53 bits, resulting in a double-rounding error.
The proper behavior is to round at fewer bits.

If you know of an efficient workaround for this problem I'd love to
hear about it!  Otherwise, I still think that you'll need to live with
some fuzz in your x86 IEEE double implementation, unless you are
willing to suffer a performance penalty.