Source code for the BNL GPU Hackathon

• https://portal.nersc.gov/project/hacc/aphearin/BNLHack19/
• https://drive.google.com/file/d/1JZn3fbBELazpnrlw-0WiGzBkNZaFlT_7/view?usp=sharing

• The code in the sf_history subpackage maps galaxy properties onto halos. For example, the mean_sfr function maps SFR onto halos across time; this function needs to be integrated to map stellar mass onto halos.
• The code in the mock_obs subpackage computes summary statistics of the galaxy population. The numba_gw_hist function calculates a Gaussian-weighted histogram, the kernel underlying a stellar mass function.

1. Install miniconda.

   I've been told it is better to keep it out of your home area, but that might just be my account. I made a directory on the global scratch area, symlinked it to my home area, and then installed there.

   The global scratch area is /hpcgpfs01/scratch.

2. Use conda init to setup your shell if you did not install it while doing miniconda.

3. Add module load "cuda/9.0" to your ~/.bashrc.

4. Install all of the software you need in your base conda env (e.g., numba, pytest, numpy, scipy, jax, etc.).

5. Get a node srun --pty -A hackathon -N 1 --exclusive -p long -t 3:00:00 /bin/bash