Xref: utzoo comp.lang.fortran:2972 sci.math.num-analysis:602
Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!usc!elroy.jpl.nasa.gov!mahendo!wlbr!voder!pyramid!weitek!weaver
From: weaver@weitek.WEITEK.COM
Newsgroups: comp.lang.fortran,sci.math.num-analysis
Subject: Re: Functions to modify IEEE floating point?
Message-ID: <22801@weitek.WEITEK.COM>
Date: 8 Mar 90 01:29:51 GMT
References: <938@orange19.qtp.ufl.edu>
Sender: weaver@weitek.WEITEK.COM
Reply-To: weaver@weitek.UUCP (Michael Weaver)
Organization: WEITEK Corp. Sunnyvale Ca.
Lines: 31

In article <938@orange19.qtp.ufl.edu> bernhold@qtp.ufl.edu (David E. Bernholdt) writes:
>
>I use Sun systems, which follows the ANSI/IEEE 754-1985 standard for
>floating point arithmetic.  Sun provides a number of functions to
>manipulate the actions of the system.  For example, [....] 
>Are these routines available on all machines which follow IEEE
>floating point?  Do they have the same names or just similar
>functionality?  

Bad news:

o IEEE-754 specifies that some way must be provided to do some of 
  these things (ieee_flags, ieee_handler, ieee_retrospective), but says 
  nothing about Fortran or what names these things will have.
  (IEEE-754 requires that underflow be gradual.) 

o To my knowlege, no standard exists on the Fortran routines for 
  control of IEEE floating point.

o Some vendors have advertised 'IEEE-format' floating point.
  This means the interpretation of the bits in a number
  follow IEEE-754, but things like rounding and exceptions
  may not.

Good News:

o IEEE-754 is short (18 pp), readable and not grossly expensive
  ($15-$20 I think). Vendor's documentation may be easier to 
  understand after reading the standard.

--Michael Weaver.