Measuring execution time in direct memory mode

[iraikov]

Hello,
I was wondering whether it might be possible to utilize ParallelContext.step_time and related functions to measure the execution time of CoreNeuron in direct memory mode. This is not a critical issue, but it helps to estimate the execution time of large network models in HPC environments with job time limits, so that the job can complete before reaching the time limit. Thanks!

[nrnhines]

For now there is only the rough timing of pc.psolve(small_tstop) using the usual interpreter timing methods. I agree that is not
too helpful to check load_balance but the issues on CoreNEURON are very different than they used to be for NEURON on uniform memory architectures. Anyway, I agree that needs to be on the agenda.