Newsgroups: comp.parallel.pvm,comp.parallel,lrz.parallel
Path: ukc!uknet!pipex!howland.reston.ans.net!gatech!hubcap!bounce-back
From: schaller@lrz-muenchen.de (Christian Schaller)
Subject: PVM on KSR computers
Message-ID: <1994Jan4.142251.1227@hubcap.clemson.edu>
Sender: fpst@hubcap.clemson.edu (Steve Stevenson)
Reply-To: schaller@lrz-muenchen.de
Organization: Leibniz-Rechenzentrum, Muenchen (Germany)
Follow-Up: comp.parallel
Date: Tue, 4 Jan 1994 14:22:51 GMT
Approved: parallel@hubcap.clemson.edu
Lines: 37
Xref: ukc comp.parallel.pvm:1129 comp.parallel:8698

We are using the Public Domain Version of PVM (3.2.4) on our
KSR for quite a while. We always had good experiences. Recently I
tried to install the ported Version of PVM 3.1, which makes use of
the Virtual Shared Memory  System. I was running in a lot of
problems. In the end I didn't release the version to the users
of our site. 

I'm writing this note to give my experiences to the public and,
maybe get some feedback from other sites running this version of
PVM. I know that this version is said to be 'not supported'.

- installing this version was rather complicated compared to 
  the Public Domain Version. I had to do some source code changes.
  Also the examples were just copied from the Public Domain Version.
  I had to insert the calls to 'ksrinit()' to allocate the 
  shared memory buffers.
- the second start of an application after starting the demon produced
  the message 'Increment MAX_SHARED_MEM and recompile'. I tried two
  work-arounds: (a) cancel the old demon and start a new one (o.k.)
  (b) change the size of MAX_SHARED_MEM and recompile the whole system.
  In the case of (b) the machine was just allocating my shared memory 
  buffer and doing nothing else. No way!
- So I always used the alternative (a). But even then the machine was not
  working as we were used to. Processes like 'ps' were not terminating
  and there was no way to kill them. We had to restart the machine.

This is some of my experiences with PVM on the KSR1. I hope this is
helpfull information for some of you. 

--
Christian Schaller                    email:   schaller@lrz-muenchen.de
Leibniz-Rechenzentrum                 phone:   +49-89-2105-8771 
Barer Str. 21                         fax:     +49-89-2809460
80333 Munich                         




