NeuroGPU + BBP TTPC Example

Prerequisites

• CUDA > 10
• NEURON > 7.6.7
• Makefile
• GCC > 8.3.0
• see requirements.txt for other python requirments

Getting Started

There are a few steps for porting NEURON models to NeuroGPU:

1. Upload hoc/mod files to a directory (like BBP_TTPC_Example folder).
2. Compile with nrnivmodl.
3. run python extractModel.pyto build a directory called PyNeuroGPU_Unix.
4. run moveFiles.sh to copy files into PyNeuroGPU_Unix.
5. chdir to PyNeuroGPU_Unix/python
6. Inside that directory you run python extractModel.py again.
7. Then run python extractModel_mappings.py.
8. Now you can run NeuroGPU by running python runBBP.py

Moving the Model

• We commonly move the simulation (NeuroGPU executable in PyNeuroGPU_Unix folder) to other repos / places for convienence.
• To do this you can just move the whole PyNeuroGPU_Unix folder to wherever you want to run the simulation.
• NeuroGPU requires that:
  • If you change the params you use allparams_from_mapping(param_values).
  • There needs to be two folders available one directory up from the directory NeuroGPU is being called from -- Data and src. Data contains files NeuroGPU needs to accesss and src contains the source code for the CUDA model.

Building Out from the Example

• To use another model one just needs to replace the hoc/mod files in BBP_TTPC_Example folder.
• Not all hoc and mod files are supported.

Please reach out to zladd@berkeley.edu or rbenshalom@ucdavis.edu if you encounter any issues. Or you can open an issue in https://github.com/roybens/NeuroGPU.

Changelog

2022-12: MOD-C-2-CPP: Compatibility fixed for NEURON 9.0, 8.2 and 8.1 2023-04: Remove generated C files and hidden files. Update branching.mod for compatibility with new data structures coming in NEURON 9.0.