search

brian-team / brian2cuda

A brian2 extension to simulate spiking neural networks on GPUs
https://brian2cuda.readthedocs.io/
GNU General Public License v3.0
61 stars 12 forks source link

Support GSL integration #242

Open denisalevi opened 3 years ago

denisalevi commented 3 years ago

Just had a quick look at brian2's GSL integration method (there is this developer documentation and here is the PR that added it to brian). Is this something that could be used in brian2cuda as well / would be beneficial? I believe there is no CUDA version of the GSL that would work in device code, but it seems an essential part of it is preparing the correct integration procedure, which could potentially be done on the CPU while performing the stateupdates on GPU? I know too little about it to tell if that would be possible and how much work it would be.

@brian-team Any opinions?

mstimberg commented 3 years ago

Sorry for replying so late, but I don't see anything to support for brian2cuda here. If there's no CUDA version of the GSL, there's not much we can do. In principle, one could run the integration on the CPU and all the rest on the GPU, but that wouldn't make much sense given that the integration is what usually benefits the most from the GPU acceleration. Also, we don't even support GSL integration with the numpy target, so this is clearly an optional feature!