home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.lang.fortran
- Path: sparky!uunet!pipex!warwick!mrccrc!doc.ic.ac.uk!agate!ames!saimiri.primate.wisc.edu!zaphod.mps.ohio-state.edu!caen!hellgate.utah.edu!lanl!cochiti.lanl.gov!jlg
- From: jlg@cochiti.lanl.gov (J. Giles)
- Subject: Re: unformatted output
- Message-ID: <1992Nov10.192736.22118@newshost.lanl.gov>
- Sender: news@newshost.lanl.gov
- Organization: Los Alamos National Laboratory
- References: <33492@adm.brl.mil> <Bx5IM3.9nF@news.cso.uiuc.edu> <Bx5s1u.C0J@pgroup.com> <KHB.92Nov3190132@chiba.Eng.Sun.COM> <18290@ksr.com>
- Date: Tue, 10 Nov 1992 19:27:36 GMT
- Lines: 18
-
- In article <18290@ksr.com>, tim@ksr.com (Tim Peters) writes:
- |> [...]
- |> Agree that an implementation is free to do less-than-perfect I/O
- |> conversions for large numbers, but the conversions have to be accurate
- |> enough in every case so that read(write(x)).eq.x always holds under round-
- |> to-nearest. From section 5.6 (Binary <-> Decimal Conversion) of 754-
- |> 1985: [...]
-
- All correct. Unfortunately, the Fortran standard does not conform to
- IEEE (in fact, it conflicts with it in some places - like with regard
- to negative zero). And the Fortran standard does not require the same
- conversion accuracy as the IEEE floating point standard. I raised
- this very point about conversion in my public review comments and
- the response was that such issues were beyond the scope of the
- committee. In other words, you're at the mercy of the implementor.
-
- --
- J. Giles
-