Distributed LHF (MPI)

To run LHF in distributed (MPI) mode use:

mpiexec --hostfile <hostfile> --map-by node -np <# processes> ./LHF -m mpi <args> -i <inputfile>

LHF needs to run in MPI mode to enable MPI to distribute the partitions and merge results. Other modes will simply repeat the same execution over multiple nodes and may have unexpected results.

--map-by node: distributes processes in a round-robin fashion

<hostfile>: MPI file that defines parameters for each node

Example:

10.0.0.1 slots=8 max-slots=8

10.0.0.2 slots=4 max-slots=4