Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watmath!clyde!rutgers!im4u!ut-sally!utah-cs!utah-gr!stride!tahoe!cecrume
From: cecrume@tahoe.UUCP
Newsgroups: comp.arch,comp.lang.fortran,comp.periphs
Subject: Implementing extended (QUAD) precision via FORTRAN on UNIX
Message-ID: <636@tahoe.UUCP>
Date: Thu, 1-Oct-87 18:29:41 EDT
Article-I.D.: tahoe.636
Posted: Thu Oct  1 18:29:41 1987
Date-Received: Sun, 4-Oct-87 06:37:03 EDT
Reply-To: cecrume@tahoe.UUCP (Charles E Crume)
Distribution: world
Organization: University of Nevada - Reno
Lines: 23
Xref: utgpu comp.arch:2321 comp.lang.fortran:213 comp.periphs:553

The University of Nevada System Computing Center (UNSCC) in Reno, Nevada;
USA needs to implement extended precision (i.e. QUAD precision) via the
FORTRAN language on a Harris HCX9 computer running UNIX.  The HCX9 will
run both BSD 4.3 and System V.

Currently, the University System uses a CDC Cyber 830 which, with its 60
bit word size, provides approximately 15 decimal digits for normal floating
point variables and 30 decimal digits for DOUBLE PRECISION variables (as
declared in FORTRAN).

As the HCX9 is a 32 bit machine, researchers here at UNR are concerned about
the reduction in precision (i.e. a DOUBLE PRECISION FORTRAN variable will
now only have 15-16 decimal digits instead of 30).

We would appreciate hearing from other people who have worked on
implementing extended precision on any 32 bit machine, especially a Harris
HCX9.  If you have (or know of someone who has) converted a FORTRAN
compiler to support REAL*16 variables, efficient algorithms to implement
it via software, some sort of add-on hardware board, etc., please contact us.

Please send your response to: Charles Crume at;

                              uunet!unrvax!tahoe!cecrume