From: Lars Rzymianowicz Newsgroups: comp.parallel.mpi Subject: Re: mpirun Date: Mon, 22 Feb 1999 17:39:58 +0100 Organization: Dept. of Computer Engineering, University of Mannheim, Germany Message-Id: <36D1885E.DE90411B@mufasa.informatik.uni-mannheim.de> References: <36CD7E76.EBE38C7@ecs.soton.ac.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Xref: ukc comp.parallel.mpi:4657 mohd ezanee rusli wrote: > Could someone tell me what is the option for mpirun that telling the program > to use different processor explicitly . Below is the example when I executed > my program and it's seem that it used the same processor. > > % mpirun -np 2 > > process 1 is alive on gemini.ecs.soton.ac.uk > process 0 is alive on gemini.ecs.soton.ac.uk Hi Mohd, there is a file called machine.ARCH in /util/machines/. It lists all possible nodes for MPI. At start-up, MPI uses a Round-Robin scheme to start the processes on the nodes. Node names can also appear more than once. Example: you have a 4-CPU SMP server and 2 1-CPU machines. Than your file could be: server server server server machine1 machine2 If you start 6 processes now, 4 are started on the SMP node, the other ones on the single-cpu machines. There is also a description in the manual.. Lars -- Homepage: http://mufasa.informatik.uni-mannheim.de/lsra/persons/lars/