From: Srinidhi <srinidhi@ece.neu.edu>
Newsgroups: comp.parallel.pvm
Subject: Re: Spawning from a spawned process
Date: Mon, 08 Feb 1999 11:15:14 -0500
Organization: Northeastern University, Boston
Message-Id: <36BF0D92.1D2A31D0@ece.neu.edu>
References: <36BB3F08.72EA71BB@ece.neu.edu> <F6uBLG.17K@liverpool.ac.uk>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Xref: ukc comp.parallel.pvm:7995


Minlin Fan wrote:

> I encountered this problem before.
>
> When a slave process is spawned, it needs some time to be ready for sending
> and receiving message. If the communications between master and slave starts
> before the slave is ready then you will get error message on "pvm_recv".
>
> There are two ways to sort this out:
> 1. do a certain time's sleep to wait for the spawned
>    slaves to be ready.
> 2. let slaves send acknowledge back to master when they are
>    ready.
>
> Minlin

Hi Minlin,

I actually tried the second way, asked the slaves to acknowledge back to the
master. Even this procedure seems very unreliable. When I tried to look at it
on "xpvm" I see that some of these slaves, sometimes, send an "mhf_invoke". Do
you or anybody know what this means? Is there a way to avoid this?

Thanks for all the help,
-Srinidhi