multithreading without the GUI

General issues of interest both for network and
individual cell parallelization.

Moderator: hines

Post Reply
staceyw
Posts: 12
Joined: Wed May 21, 2008 3:49 pm

multithreading without the GUI

Post by staceyw »

I have a working model, written to be MPI-ready, that was running on 6.1. What needs to be done to run this with multithreading capabilities in 7.0? I do not use the GUI, and run long batches, so it needs to be either commandline changes or something to write into the .hoc files (sorry, I haven't been able to find this info, though I know it must be here somewhere).
ted
Site Admin
Posts: 6289
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: multithreading without the GUI

Post by ted »

First question is whether multithreaded execution will help. Do you have cells with lots of compartments and mechanisms? i.e. requiring at least a couple thousand ODEs per cell? If no, multithreaded won't help. If yes, the next thing is to make sure that it's compatible with multithreaded execution--see
http://www.neuron.yale.edu/neuron/stati ... ml#Threads, and also see
http://www.neuron.yale.edu/phpBB/viewto ... =22&t=1476
especially with regard to using mkthreadsafe to check and fix mod files.

For any given model, discovering if multithreaded execution is beneficial, and how many threads to use, is an empirical question. It's probably a good idea to use the GUI--specifically the RunControl panel and the NEURON Main Menu / Tools / Parallel Computing tool--to run a few interactive tests in order to decide on appropriate simulation parameters.
Post Reply