Newsgroups: comp.parallel.pvm
From: Anatol Piotrowski <piotrows@scs.carleton.ca>
Subject: Warning: Bug (?) with pvm_bcast ?
Organization: Carleton University
Date: Tue, 11 Jun 1996 14:52:17 -0400
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Message-ID: <31BDC061.5A2C@scs.carleton.ca>

Following situation:
	Master and slave (just one) run on the same machine. Master bcasts to slaves using PvmDataInPlace. 
Slave receives using pvm_precv. 1024 floats are sent. (no problem for 961 floats). All machines are pentiums.
	This fails when master and slave both run on the machine which is not the machine on which PVM was 
started on. (slave receives corrupted data).
	If master/slave is run on the same machine that pvm was started on messages are not corrupted.

I fixed the problem by using PvmDataRaw instead of PvmDataInPlace.
I checked I don't modify data so PvmDataInPlace should work.

