Author: Mark Plagge, Rensselear Polytechnic Institute Copyright: 2019 - MIT LICENCE Version 2.0
NeMo is a simulation model designed to allow experimentation with different neuromorphic hardware. Built on top of ROSS, NeMo features excellent scalability across HPC clusters. NeMo is targeted at current neuromorphic hardware designs, and as such currently simulates spiking neural networks in a grid layout, similar to the IBM TrueNorth Processor.
NeMo has undergone significant changes since the initial release. As such, NeMo v2 and later are now contained in the NeMo2 repository, while the original NeMo code base is in the NeMo repo.
Benchmarks and results of the first edition of NeMo is available at the ACM Digital Library.
Doxygen documentation is available as well.
Install instructions are coming soon, and will be placed on the wiki.
Some notes:
Config BGRS for OMP Target Offload:
openmp offload BGRS:
Configure /w CUDA 9.2 & XLC
xlc_configure -o config_xlc -cuda $CUDA_PATH
xlc++ -F config_xlc -qsmp=omp -qoffload
Possibly the same for mpicc
XLC++ expects CUDA 9.2 (XLC 16.0.1)
*Code Project & Statistics*