Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!columbia!caip!clyde!cbatt!cbosgd!ihnp4!inuxc!pur-ee!uiucdcs!convex!wallach
From: wallach@convex.UUCP
Newsgroups: net.arch
Subject: Re: Long integers on the Cray?
Message-ID: <82200003@convex>
Date: Mon, 11-Aug-86 14:26:00 EDT
Article-I.D.: convex.82200003
Posted: Mon Aug 11 14:26:00 1986
Date-Received: Wed, 13-Aug-86 05:37:38 EDT
References: <1068@jade.BERKELEY.EDU>
Lines: 19
Nf-ID: #R:jade.BERKELEY.EDU:1068:convex:82200003:000:634
Nf-From: convex.UUCP!wallach    Aug 11 13:26:00 1986



ne way to do this on a cray is the following.

partition the long integer into 63  (yes 63) bit partions. the higher
order bit is always set to 0. in th vector accumulator.

perform the add, etc.  compare the output with positive or negative.
(there is no compare vector on the cray) so use their test vm stuff.

then expand vm into a vector of integer 0 or 1.  this is the carry
out.  then add the carry vector and so on.

you want to check with some of the people as lawrence livermore.

their code is in the public domain.  call frank McMann in the compuer
science group.

steve wallach (convex computer .... convex!wallach)