How to run hedge codes in parallel

Almost any code in hedge can be run in parallel. Exceptions (as of 2011-05-05) include codes that use quadrature and second-order operators. (Both of those aren't hard restrictions to lift, it's simply not quite done yet.) In particular, wave equations, Maxwell's, Euler's (non-quadrature) work fine.

To run with CUDA, simply start

$ python DRIVER.py -f cuda

To run with MPI, do

$ mpirun -np 17 python DRIVER.py -f mpi

To run with both, do

$ mpirun -np 17 python DRIVER.py -f mpi,cuda

Hedge/HowTo/RunningInParallel (last edited 2013-01-16 23:25:32 by AndreasKloeckner)