Optimizing performance

Running simulations outside the wrapper

To run simulations outside the MATLAB/Python environment, e.g. when computing on an HPC cluster: run first the wrapper only to generate the input file qdyn.in, then run the qdyn executable outside the wrapper.

Managing parallel computing

OpenMP

For 3D simulations on 2D faults (MESHDIM=2 or 4), QDYN is parallelized for shared memory multi-processor systems with OpenMP. Before compiling the code, you should set the specific compiler optimisation flags that enable OpenMP, as described in the Makefile (see step 2.b in section 2.3). Before performing parallel simulations, you should set the following environment variable:

setenv OMP_NUM_THREADS 8

This command allows QDYN to run on 8 threads, which will roughly speed up calculations by a factor of 8. The number of threads should be set according to demand. In general, set this value to the maximum number of threads available on your shared memory system.

MPI

For 3D simulations with MESHDIM=2, QDYN can run in parallel in distributed memory clusters with MPI. The number of processors must be set in the variable NPROCS.