The next step is the description of the VPM, which will execute mpC applications. The description is provided by a manually written VPM description file, which should be in the $MPCTOPO directory. The name of this file is just considered as the name of the specified VPM.

D.6.1. VPM Description File

A VPM description file consists of lines of three kinds. Lines beginning with the symbol # are treated as comments. Lines containing one of the symbols {or} are used to specify the hierarchy of clusterization. All other lines should be in the following format:

<name> <number_of_processes> <number_of_processors>(opt) where
  • <name> is the name of the computer as it appears in the system /ets/hosts file

  • <number_of_processes> is the number of processes of the computing space that will be running on this computer, and

  • <number_of_processors> is a bracketed integer constant which specifies the number of physical processors of the SMP computer.

While the mpC programming environment can automatically determine the number of processors of each computer during creation of the VPM, the user may specify the number of processors explicitly. The host workstation must go first in the file. The abstract host-processor will be mapped to a process running on this workstation.

For example, the following file describes a VPM that consists of three workstations (alpha, beta, and gamma) each running five processes, and the host workstation is alpha:

# three workstation each running 5 processes alpha 5 beta 5 gamma 5

The following example describes a VPM with the same total number of the processes, but running on a single workstation. It may be useful for debugging mpC applications:

# simple VPM for debugging alpha 15

The following more complicated example describes a VPM with clusterization:

# four 1-processor workstations... omega 1 #...including cluster: {    alpha 5    beta 5    gamma 5 } # and one 2-processor server serv 10 [2]

Note that the actual total number of running processes is greater than the number specified in the description file. A process for the dispatcher is added automatically and runs on the host workstation. The mpC abstract host-processor is always mapped to the host workstation.

Parallel Computing on Heterogeneous Networks
Parallel Computing on Heterogeneous Networks (Wiley Series on Parallel and Distributed Computing)
Year: 2005
Pages: 95

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net