See !16 for detailed output of running parallel-neuro-sim with different numbers of threads for a simple neuron config. It appears that OpenMP is adding significant overhead slowing down the program's overall run time. Since our LUT hardware will not allow parallelization (as of right now) it might be best to never build with the -DUSE_OPENMP=ON option.
In GitLab by @heidecjj on Jan 3, 2019, 15:01
See !16 for detailed output of running parallel-neuro-sim with different numbers of threads for a simple neuron config. It appears that OpenMP is adding significant overhead slowing down the program's overall run time. Since our LUT hardware will not allow parallelization (as of right now) it might be best to never build with the
-DUSE_OPENMP=ONoption.