• Linux distribution with wget and git installed
  • Nvidia GPU card with CUDA toolkit >= 10.1 (Older versions could be available on request)

Installation of Anaconda/Miniconda

COMPASS binaries, which contain the optimized GPU code, can be installed via Anaconda. Then, you have to install Anaconda 3 or Miniconda 3 (python 3 is required).

We recommend Miniconda instead of Anaconda as it is much lighter, but it’s up to you.

export CONDA_ROOT=$HOME/miniconda3
bash -b -p $CONDA_ROOT

Don’t forget to add your Miniconda or Anaconda directory to your PATH:


Installation of COMPASS via conda

Once Miniconda is installed, installing the COMPASS binaries is easy :

conda install -c compass compass -y

Note: conda main channel is compiled with CUDA 11.6.0, for previous version please post an issue and will provide it.

This command line will also install dependencies in your conda environment.

Installation of SHESHA package for COMPASS

First, you will need to set some environment variables:

export SHESHA_ROOT=$HOME/shesha

Finally, you can get the Shesha package of COMPASS. This python package is the user level of COMPASS. It also contains all the initialization functions.

git clone $SHESHA_ROOT

Test your installation

Once the installation is complete, verify that everything is working fine :

cd $SHESHA_ROOT/tests

This test will basically launch fast simulation test cases and it will print if those cases have been correctly initialised.

Run the simulation

You are ready ! You can try it with one of our paramaters file:

ipython -i shesha/scripts/ data/par/par4bench/

And if you want to launch the GUI:

ipython -i shesha/widgets/

Browse through the manual to understand how to use COMPASS.