Newsgroups: comp.parallel.pvm
From: wrankin@duke.edu (William T. Rankin)
Subject: Threads and PVM 3.4beta?
Organization: Duke University EE Dept.; Durham, NC
Date: 30 Jul 1997 14:38:32 GMT
Message-ID: <5rnjl8$5fi$1@news.duke.edu>


Firstoff, thanks to the PVM group for all their efforts in brining out
3.4.  I am downloading it now and will try to give some feedback on any
problems I run into, as well as the new features (I am *really* looking
forward to using the message-handler features :-)

(And now that I've buttered them up enough... ;-)

The PVM homepage advertized that thread safty issues would be addressed
in 3.4, but I noticed that they are not on the 3.4-beta3 brag sheet.
Is there any update on when/if we can expect a thread safe PVM 3.4?

In the meantime, will the following workarounds "work"?

  - I assume that I can have PVM in a multi-threaded program as long as
    all PVM calls are performed by the same thread.

  - I can take the approach of Adam Ferrari up at Virginia (is
    Adam still there?) w/ TPVM and wrap all my PVM calls with mutex
    locks.  However, can I have *one* simultanious pvm_send()  with
    a pvm_recv()? ..  or must the receive and send calls also be
    mutually exclusive with each other?

Any other suggestions would be appreciated.

Thanks for all the help,

-bill rankin

-- 
bill rankin ...................................... philosopher/coffee-drinker
wrankin@ee.duke.edu ........................................ doctoral wannabe
duke university dept. of electrical engr ......... scientific computing group