home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!paladin.american.edu!auvm!VNET.IBM.COM!MFC
- Message-ID: <REXXLIST%92122820260479@UGA.CC.UGA.EDU>
- Newsgroups: comp.lang.rexx
- Date: Mon, 28 Dec 1992 14:33:01 GMT
- Sender: REXX Programming discussion list <REXXLIST@UGA.BITNET>
- From: Mike Cowlishaw <mfc@VNET.IBM.COM>
- Subject: Even/Odd rounding
- Comments: To: rexxlist@uga.uga.edu
- Lines: 15
-
- (Re: comment by J. Brock -- apologies for delay.)
-
- Well, the problem goes a lot further than simply treating overprecise
- input numbers: once one decides to use Even/Odd rounding, then one must
- use it consistently: it should be used for all calculations and rounding
- (or else you'll get different results depending on the source of a
- number). If Even/Odd rounding is the rule for arithmetic, then
- calculations are required to use all digits in their inputs (in order
- to provide a result of maximum possible length, that can then be
- Even/Odd-rounded). Hence all calculations have to be done to arbitrary
- lengths: doable, of course, but at considerable cost for a very tiny
- benefit. If you have a critical application where the rounding method
- has a real effect, it's much better to increase NUMERIC DIGITS.
-
- Mike Cowlishaw
-