Example request

[TeroFrondelius]

I saw your JuliaCon talk today. Impressive work. README says:

Using hybrid parallelism, where threads are used to communicate within each machine and MPI or Julia's native parallelism between machines is much faster. For MPI, the scaling is almost ideal with the number of machines, but for DArray the results are more erratic.

Would you mind making an example to show how this is done?

[barche]

The example is actually here: https://github.com/barche/MPIArrays.jl/blob/master/test/matmul_mpi.jl

You need to set OPENBLAS_NUM_THREADS to choose the number of threads.