> It seems that the only way to have a no-shared memory > communication-only based parallel programming model (la MPI) is by > using the spmd construct and using labindex to customize code > execution. Is there a way for the main matlab process to communicate > with the workers in spmd while they are working?
There is not. The MATLAB client does not in fact have an MPI connection to the workers.
> The way I've been getting around this is by assigning one worker as a > 'master' and others as 'workers'