11. Parallel CommandsΒΆ

The parallel commands are currently only working in the Linux version. The parallel OpenSeesPy is similar to OpenSeesMP, which requires users to divide the model to distributed processors.

You can still run the single-processor version as before. To run the parallel version, you have to install a MPI implementation, such as mpich. Then call your python scripts in the command line

mpiexec -np np python filename.py

where np is the number of processors to be used, python is the python interpreter, and filename.py is the script name.

Inside the script, OpenSeesPy is still imported as

import openseespy.opensees as ops

Common problems:

  1. Unmatch send/recv will cause deadlock.

  2. Writing to the same files at the same from different processors will cause race conditions.

  3. Poor model decomposition will cause load imbalance problem.

Following are commands related to parallel computing:

  1. getPID command

  2. getNP command

  3. barrier command

  4. send command

  5. recv command

  6. Bcast command

  7. setStartNodeTag command

  8. domainChange command

  9. Parallel Plain Numberer

  10. Parallel RCM Numberer

  11. MUMPS Solver

  12. Parallel DisplacementControl

  13. partition command