The harmonic balance method procedure automatically partitions the model during solution to run in distributed-memory parallel (DMP) mode using two processors. The nonlinear multiharmonic solution runs on one processor, the master MPI process, while the AFT transient analysis runs on the other processor. Consequently, the minimum number of MPI processors is 2, and if more processors are requested, they are terminated.
Hybrid-parallel processing is supported, allowing the multiharmonic solution process to run in shared-memory parallel (SMP) / multithreading. The figure below shows a diagram depicting the processing distribution particular to HBM. Multithreading in HBM is efficient for systems with a larger number of linear degrees of freedom and/or harmonics.
To activate hybrid-parallel processing, you must issue the -nt command line option to specify how many threads should be used for the multiharmonic MPI process (see Using Hybrid Parallel Processing (SMP + DMP) and Specifying Processes, Threads, and Cores in SMP, DMP, and Hybrid Parallel in the Parallel Processing Guide). An HBM analysis does not automatically activate hybrid-parallel processing.
Note: HBM simulations that include a modal analysis or substructuring generation pass can use hybrid-parallel processing only if the HBM solving is done in a separate Mechanical APDL run.