Dear NEURON users,
I've written a large IF network simulation in MATLAB. It simulates a dissociated cortical network on a multi-electrode array (stimulation/recording sites). I'm looking to increase the computation speed and have a cluster at my disposal. I was wondering if anyone thinks this would be something I should try to write in NEURON given having no prior experience with it, or if there are other programs out there I should try (CSIM, GENESIS?).
Thanks!
N
should I port my sim?
Moderator: wwlytton
-
ted
- Site Admin
- Posts: 6299
- Joined: Wed May 18, 2005 4:50 pm
- Location: Yale University School of Medicine
- Contact:
Re: should I port my sim?
First decide whether it's worth the effort. Is T_S > T_L + T_P + T_S* ?
T_S = time required to run all the simulations you want with your existing software on existing hardware
T_L = time required to learn a new simulation environment
T_P = time required to port your model to that environment = time to write new code + time to debug new code, where time to debug new code = (time to perform test simulations + compare them with control simulation results generated with your existing code + time to figure out why the results aren't identical + time to make the necessary changes so that they are identical)
T_S* = time required to run all the simulations you want with your new software on new hardware
Unless T_S is very large, and T_S* is much smaller, it's probably not worth the effort. T_L can be quite large if you are starting from "no knowledge" about the new simulation environment. That will also make T_P large. And T_P will be even larger if your original model uses "idioms" that do not readily map into the new simulation environment. You might be able to save some time by teaming up with someone who is very familiar with the new simulation environment.
The following is relevant only if you have not already decided to give up on the whole idea.
This article, which is available from http://www.neuron.yale.edu/neuron/bib/nrnpubs.html, presents an informed survey of simulators that are capable of running on parallel hardware and is still quite current:
Brette, R., Rudolph, M., Carnevale, T., Hines, M., Beeman, D., Bower, J.M., Diesmann, M., Goodman, P.H., Harris, F.C.J., Zirpe, M., Natschläger, T., Pecevski, D., Ermentrout, B., Djurfeldt, M., Lansner, A., Rochel, O., Vieville, T., Muller, E., Davison, A., El Boustani, S., and Destexhe, A. Simulation of networks of spiking neurons: a review of tools and strategies. J. Comput. Neurosci. 23:349-398, 2007.
T_S = time required to run all the simulations you want with your existing software on existing hardware
T_L = time required to learn a new simulation environment
T_P = time required to port your model to that environment = time to write new code + time to debug new code, where time to debug new code = (time to perform test simulations + compare them with control simulation results generated with your existing code + time to figure out why the results aren't identical + time to make the necessary changes so that they are identical)
T_S* = time required to run all the simulations you want with your new software on new hardware
Unless T_S is very large, and T_S* is much smaller, it's probably not worth the effort. T_L can be quite large if you are starting from "no knowledge" about the new simulation environment. That will also make T_P large. And T_P will be even larger if your original model uses "idioms" that do not readily map into the new simulation environment. You might be able to save some time by teaming up with someone who is very familiar with the new simulation environment.
The following is relevant only if you have not already decided to give up on the whole idea.
This article, which is available from http://www.neuron.yale.edu/neuron/bib/nrnpubs.html, presents an informed survey of simulators that are capable of running on parallel hardware and is still quite current:
Brette, R., Rudolph, M., Carnevale, T., Hines, M., Beeman, D., Bower, J.M., Diesmann, M., Goodman, P.H., Harris, F.C.J., Zirpe, M., Natschläger, T., Pecevski, D., Ermentrout, B., Djurfeldt, M., Lansner, A., Rochel, O., Vieville, T., Muller, E., Davison, A., El Boustani, S., and Destexhe, A. Simulation of networks of spiking neurons: a review of tools and strategies. J. Comput. Neurosci. 23:349-398, 2007.