Newsgroups: comp.parallel.pvm From: Matt Beare Subject: PvmRouteDirect Organization: University of East Anglia Date: Wed, 20 May 1998 12:44:39 +0100 Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-ID: <3562C227.167E@uea.ac.uk> On a workstation cluster (DEC Alphas) it was suggested to me that I use the following command at the start of my program to speed up communications: pvmfsetopt (PvmRoute, PvmRouteDirect, OldVal) I gather that this allows message-passing to by-pass the PVMD, using direct TCP sockets. Sure enough communication is seen to be significantly faster, but from time to time my application will hang, for no apparent reason, whereas when using the slower method of going via the PVMD, the application is perfectly stable. Can anyone suggest why this is. Papers that I have read suggest the opposite should be true and that the PvmRouteDirect option should be more stable. My question then, is if it is more stable and significantly faster, why is PVMD routing ever used ? Thanks, Matt. -- ----------------------------------------------------------------- Matthew Beare | Email: M.Beare@uea.ac.uk School of Mathematics | Tel: +44 (0)1603 592990 University of East Anglia | Fax: +44 (0)1603 259515 Norwich NR4 7TJ England | http://www.mth.uea.ac.uk/~h007/ -----------------------------------------------------------------