Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.csd.uwm.edu!bionet!agate!shelby!unix!garth!fouts@garth.ingr.com (Martin Fouts)
From: fouts@garth.ingr.com (Martin Fouts)
Newsgroups: comp.unix.cray
Subject: Re: C++ on cray UNICOS??
Message-ID: <3210@garth.UUCP>
Date: 17 Aug 89 16:26:18 GMT
References: <241@uw-apl.UUCP> <3930@hall.cray.com>
Sender: fouts@garth.UUCP
Distribution: na
Organization: INTERGRAPH (APD) -- Palo Alto, CA
Lines: 60
In-reply-to: rosenkra@hall.cray.com's message of 1 Aug 89 01:45:58 GMT


In article <3930@hall.cray.com> rosenkra@hall.cray.com (Bill Rosenkranz) writes:

   In article <241@uw-apl.UUCP> keith@uw-apl.UUCP (Keith Kerr) writes:
   =Is there an implementation of C++ running on cray./UNICOS
   =systems??  If not, could anyone venture a guess as to
   =how tough a port would be.  I've heard that UNICOS was
   =derived from some unix version, but I don't know anything
   =more about it.


   UNICOS is based on UNIX System V.3 so in theory, cfront should be possible.

   i am not an expert on c++, but as i understand it, the research version
   (1.2.1?) is a preprocessor which generates c code. cray has an ANSI-
   compliant C as well as a vectorizing C which so far i have been able
   to port numerous codes with (i have just started a port of GNU emacs
   18.54 for a cray-2 under UNICOS 5.0, for example). so c++ is text in,
   text out and it should not be a problem. the question is whether the
   c code generated can readily be compiled. i suspect it can.

   i have no knowledge of cray's plans for c++ but i suspect somebody
   is playing around with it :^).

   -bill
   rosenkra@boston.cray.com
   [i speak for me, not cray, so don't quote me...]


Cfront should be easy to port, barring an potential number of word
size and alignment problems.  The problem isn't the preprocessor, it
is the library support.  This would be much more interesting.  If
you've got experience providing run time libraries and you understand
UniCos, the calling sequence, and CAL, it should be a straightforward
but time consuming effort.

The other alternative is G++, which requires Gnu C.  I started a Gnu C
port to the Cray 2, and it looks like a lot of work because of the way
in which Crays differ from the model provided by the Gnu compiler, so
I would recommend the first approach.

Marty


--
Martin Fouts

 UUCP:  ...!pyramid!garth!fouts  ARPA:  garth!fouts@ingr.com
PHONE:  (415) 852-2310            FAX:  (415) 856-9224
 MAIL:  2400 Geng Road, Palo Alto, CA, 94303

I'll say something clever, as soon as I get some caffeine in my blood stream
-- 
Martin Fouts

 UUCP:  ...!pyramid!garth!fouts  ARPA:  garth!fouts@ingr.com
PHONE:  (415) 852-2310            FAX:  (415) 856-9224
 MAIL:  2400 Geng Road, Palo Alto, CA, 94303

I'll say something clever, as soon as I get some caffeine in my blood stream