From rbk@usl.edu Tue May 6 17:29:40 1997 Received: from bp.ucs.usl.edu (root@bp.ucs.usl.edu [130.70.40.36]) by dkuug.dk (8.6.12/8.6.12) with SMTP id RAA18620 for ; Tue, 6 May 1997 17:29:10 +0200 Received: from rbk5287.usl.edu (liberty.usl.edu) (liberty.usl.edu [130.70.46.171]) by bp.ucs.usl.edu with SMTP id AA05951 (5.65c/IDA-1.4.4 for ); Tue, 6 May 1997 10:28:29 -0500 Message-Id: <2.2.32.19970506152740.00673974@pop.usl.edu> X-Sender: rbk5287@pop.usl.edu X-Mailer: Windows Eudora Pro Version 2.2 (32) Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Date: Tue, 06 May 1997 10:27:40 -0500 To: Michael Metcalf , sc22wg5@dkuug.dk From: "R. Baker Kearfott" Subject: Re: (SC22WG5.1370) Floating-point arithmetic At 04:09 PM 5/6/97 +0200, Michael Metcalf wrote: > >There is a fascinating article, "On the need for predictable >floating-point arithmetic in the programmming languages Fortran 90 and >C/C++" in the March issue of SIGPLAN. It criticises both the language >standards and their (named) implementations (although C++ comes off better >than f90!). It applauds John Reid's efforts to improve the situation. I also applaud John Reid's efforts. > >Full details are at ftp://hwins.uia.ac.be/pub/CANT/SIGPLAN. Hopefully I'll be able to get this when there is less network traffic :-( > >My conclusion would be that, since this appears to be an area where C++ >beats f90 on its home ground, getting this right in F2000 deserves a >higher priority than burdening the language with interval arithmetic. My personal view is that predictable floating point and interval arithmetic are, at the most basic implementation level, two facets of the same thing. In particular, if the accuracy (distance from the exact result) could always be known from Fortran in an efficient and portable way, then the most basic operations of interval arithmetic could be implemented relatively efficiently (although not as well as an intrinsic implementation) within Fortran. (Here, I include base conversions and I/O when I say "floating point operation.") In principle, an interval standard function library could also be built on top of that, although such a library may not take full advantage of a vendor's high-quality floating point intrinsic library. The remaining issues of interval arithmetic are those of syntax and semantics, dealing with both arithmetic operations and operations such as intersection and union. We want to give a vendor the leeway to develop the most efficient possible implementation with the most natural possible user interface. On the other hand, it would be a shame to see more than one successful implementation with incompatible syntax. Best regards, --------------------------------------------------------------- R. Baker Kearfott, rbk@usl.edu (318) 482-5346 (fax) (318) 482-5270 (work) (318) 981-9744 (home) URL: http://interval.usl.edu/kearfott.html Department of Mathematics, University of Southwestern Louisiana USL Box 4-1010, Lafayette, LA 70504-1010, USA ---------------------------------------------------------------