www.neuron.yale.edu

Hi
I have what seems to be a weird question. Usually when I want to do multisplit on a model that I am running. I do:
Tools-> Parallel Computing -> threads=8 & Multiplit on
Refresh

and it works. But I can't seems to find the correct commands so that this clicks will be automatically run from the code,
When I do the following code:

pc.nthread(8)
pc.multisplit()

I do not see any improvement in the speed of the simulation. Only when I click on the gui on Multisplit checkbox I then see any effect on the simulation time..
I understand that there need to be some kind of load balance command, but I can't seem to find it.

How can I activate the gui Multisplit() command via code?

Thank You

The simplest way to proceed is to first save a configured ParallelComputeTool to a session file. Then you can load that session file and you'll get multithreaded execution, but the only line of code you have to write is a load_file() or xopen() statement that reads the ses file.

"Oh, but I might want to change the number of threads."

Easy. After you load the ses file, there will be an instance of the ParallelComputeTool class called ParallelComputeTool[0]. Just do this: You can also turn multsplit on and off--just substitute
multisplit(1) // 0 to turn off
for the change_nthread() statement.

Be sure to read the Programmer's Reference about object_push() and object_pop(). Also examine the ParallelComputeTool's .ses file, and look in the hoc library at the contents of parcom.hoc and maybe you'll get some new ideas.

Thank You,
Working Great!

Write as little code as necessary. Saves time and effort, avoids mistakes.