How to implement MPI_ALLREDUCE in AMREX?

[solarmaterial]

Hey AMReX team,
When dealing with parallel computation, each processor computes a variable, e.g., the time step dt_Local. However, computaion on all the processors should advance with the same time step, therefore, we need the minimum dt_Local on all the processors. This can be realized with the MPI_ALLREDUCE, but how to implement in AMREX?

Looking forward to your reply!

Yours sincerely,
solar

[atmyers]

Hi solar,

AMReX_ParallelDescriptor.H contains a number of wrappers around MPI functions like that, including for example ReduceRealMin, which can be used to do what you want here.

-Andrew

[solarmaterial]

Thank you for your replies! I'll try it.