HOW TO RUN
----------
To compile and run the code on Euler remove all previously loaded nodules (purge) and load the following modules:

module purge
module load new
module load gcc/6.3.0 
module load python/3.7.1

or run `source modules.src`
Make sure that you loaded this files during compilation of Korali.


Request interactive shell
-------------------------

bsub -n 4 -R -W  04:00 -Is bash

and compile with `make`, respectively `make ssa` or `make cmaes` for partial builds.
You may also run with more or less nodes.

Run
-------------------------

export OMP_NUM_THREADS=4 
./ssa
./cmaes

You may want to set the OMP_NUM_THREADS variables to different values.

If interested feel free to use input arguments and study the behaviour of the system for different inputs, e.g.
./ssa -omega 1 -samples 5000
Lower quantities usually increase the stochastic effects and require more simulations/samples.


WHAT TO MODIFY
--------------
You only need to change the indicated sections in 
- SSA_CPU.cpp
- includes/objective.hpp

and 
- main_cmaes.cpp
if you wish to modify the population size of CMA-ES.


CODE OUTPUT 
-----------
When the ./ssa code runs, it prints

the averaged time for 1 SSA simulation, the FLOPs, Byte Transfers, and the performance.
Note that the latter values depend on your implementation of SSA::getTransfers() and SSA::getFlops() and are wrong estimates if not (correctly) implemented.