Newsgroups: comp.parallel.pvm
From: wrankin@ee.duke.edu (William T. Rankin)
Subject: Re: PVM & Signals
Keywords: signals, pvm, tasks
Organization: Duke University EE Dept.; Durham, NC
Date: 12 Feb 1995 19:23:56 GMT
Message-ID: <3hln8c$oj9@news.duke.edu>

In article <3hjtee$48t@news.cs.tu-berlin.de>, demos@cs.tu-berlin.de (Alexander Weidt) writes:
|> sth@em.informatik.uni-frankfurt.de (Stefan Hoereth) writes:
|> 
|> >  is there a way to send a signal to a task when
|> >a new message for it has arrived (or is there just
|> >message 'polling') ?

yes.

you can use pvm_signal() from the sending task to "get the attention"
of the receiving task so that it can go receive the message.  you
need to be *VERY* careful about masking to prevent a sending process
from interrupting the receiving task while it is in the middle of 
any other PVM operation.  this is a non-trivial programming problem at best.

Kind of a "Poor man's Active Message"

|> The bad news is that PVM is sorely lacking such asynchronous
|> message arrival notification.
|> 
|> Best regards, Alex.

Well, Alex, given that writing re-entrant code for message passing is
(to say the least) a royal pain to both program and (especially) debug,
i personally agree with the decision not to implement it within the 
current version.

Maybe in the future. :-)

-- 
----                                /       __/    /    /
bill rankin                        /              /    /
wrankin@ee.duke.edu               ___  /    /    /    /
philosopher/coffee-drinker       /    /    /    /    /
molecules-R-us                  /    /    /    /    /
                             _______/  __/  __/  __/

We staunch traditionalists know that appearence is everything.
Technique is nowhere near as important as having your pleats straight
when you die
                                 - Steve Gombosi


