Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!rutgers!dayton!rosevax!umnd-cs!max
From: max@umnd-cs.UUCP (Max Benson)
Newsgroups: comp.lang.c,comp.unix.questions,sci.math
Subject: Re: Wanted: arbitrary precision integer arithmetic
Message-ID: <266@umnd-cs.UUCP>
Date: Thu, 4-Dec-86 12:25:47 EST
Article-I.D.: umnd-cs.266
Posted: Thu Dec  4 12:25:47 1986
Date-Received: Fri, 5-Dec-86 05:15:24 EST
References: <4056@jhunix.UUCP> <556@maynard.UUCP>
Distribution: na
Organization: University of Minnesota at Duluth, Computer Science
Lines: 20
Keywords: arbitrary precision integer arithmetic
Summary: Bugs in UNIX mp library, have alternative
Xref: mnetor comp.lang.c:243 comp.unix.questions:249 sci.math:277

Last time I checked there remain serious bugs in the UNIX mp
library that remain from the PDP-11 days.  One was in the divide
routine that made it divide by zero for only certain large numbers.
This was because the exact integer representation wasn't normalized
at some point.  Another problem was that the in the add/subtract
routine it gets signs mixed up in one case.  I guess that the UNIX
programs that use mp don't ever find the add/subtract bug because
they remove the signs before getting to mp, but I imagine that the
divide bug is out there waiting for you in dc.  I found and tracked
down these bugs for my own purposes, but now I have written a new
library (in C, hopefully vanilla, currently runs on SUN, UMAX4.2) of
arbitrary precision integer arithmetic that is somewhat faster and
better yet public domain!  I can share this with those
interested.  It is part of a larger library of routines for symbolic
mathematics which will be eventually freely available once it is
documented.

Max Benson:  University of Minnesota-Duluth
max@umn-duluth.csnet
...!ihnp4!umn-cs!umd-cs!max