Over-threading on HPC.

[augeas]

Hello, a user of our HPC system has encountered over-threading issues, and as this software is their script's only external dependency, and makes no other use of multiprocessing, it is highly likely to be the cause. multiprocessing.cpu_count(), as I have frequently noticed, does not return the correct value on HPC schedulers. If the desired number of cores is not to be passed in explicitly, it could be read from OMP_NUM_THREADS, which is reasonably idiomatic, as numpy.linalg does the same.

In AbstractSustain.py:

self.pool  = pathos.multiprocessing.ProcessingPool() #pathos.multiprocessing.ParallelPool()
self.pool.ncpus  = multiprocessing.cpu_count()

might become:

self.pool = pathos.multiprocessing.ProcessingPool(
    int(os.environ.get('OMP_NUM_THREADS', multiprocessing.cpu_count()))
)