MPI programs can be run using a mixed configuration of nodes and number of processes per node. Please, refer to your MPI local implementation for further details.

 
$  mpirun -np [number_of_processes] /xmipp/install/dir/bin/xmipp_mpi_mlf_align2d ...

The parameters used are those of the sequential version, therefore see the Mlf_align2d manual page for more information. Actually, the parallelized version should give exactly the same results as the sequential one (if no random functions are used).

Parallelization is performed by splitting the input selfile of experimental images in np equally-sized subsets. Each of the processors will process one of these subsets. Once all processors have processed their subset, the results are combined and the next iteration is started. (Caution: this means that slower (or more occupied) processors may slow down the entire process...) You may select which nodes to run on providing-machinefile file.dat, after the-np [number_of_procs] option. Here,[[FileDat]] is a file containing a list of those nodes you wish to run on.

--Main.SjorsScheres - 10 Sep 2007