MATLAB: How can i implement multilevel parallelism on a parallel cluster distribute shared-memory

MATLAB Parallel ServerParallel Computing Toolbox

i am using a HPC architecture that it's made of 8 nodes, each blade consists of 2 quadcore processors sharing memory for a total of 8 cores and of 64 cores, in total. how can I set up a cluster profile to run my code on p (1=<p<=8) nodes? and fixed a node, how can I set up another cluster profile to run my (sequential) code on c (1=<c<=8) cores ? how many workers should I consider in both cases?

thanks

Best Answer

MATLAB does not support multi-level parallelism with the Parallel Computing Toolbox.

What it does support is allocating more than one thread (core) to a parallel worker. These threads can be used by MKL or LAPACK high performance libraries that automatically act in parallel without starting new processes.

You cannot, however, have a parallel worker run a parfor in order to have workers for workers.

Related Solutions

MATLAB: Calling parpool with SpmdEnabled = False

Unfortunately, only MJS and Local cluster types support SpmdEnabled = false. You might be able to use the "cluster parfor" approach though - see the documentation. Basically, you would transform your main parfor loop like so:

% Important: do *not* create a parallel pool prior to running this!
% In fact, you may wish to call "delete(gcp('nocreate'))" to be sure.
% Get the cluster object
c = parcluster('HPCServerProfile1');
% Create the parforOptions structure
opts = parforOptions(c);
% Run the parfor loop directly on the cluster with no parallel pool
parfor (idx = 1:10000, opts)
    % Body of the parfor loop
end

This should be more robust, however it does incur more overhead than the interactive parpool approach.

MATLAB: MATLAB Distributed Computing Server

Yes, you can open a matlabpool using MDCS workers and run PARFOR and SPMD.

Best Answer

Related Solutions

MATLAB: Calling parpool with SpmdEnabled = False

MATLAB: MATLAB Distributed Computing Server

Related Question