brian-team / brian2cuda

A brian2 extension to simulate spiking neural networks on GPUs

Home Page:https://brian2cuda.readthedocs.io/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Report when models are using too many registers for full occupancy

denisalevi opened this issue · comments

See #266. I would like to have this information included in the INFO prints.

  • What is the maximal active block size per SM.
  • If it es less then given by the maximal threads per SM, mention that.
  • If the number of threads is reduced to allow more blocks / SM, mention that.

If there are measures to reduce this, a user warning should be printed. But that would need some form of C++/CUDA logging system, #103.