Installing NCL on greatlakes (03/22/2023)

Installing via Conda

1. Install conda:

• Follow this link and download the Linux 64-bit (that is, x86_64) installer script.
• Run bash ${FILE_NAME} substituting the name of the downloaded script (it begins with Miniconda if you want to use tab completion).
• Follow the prompts. Ensure that the conda root is located on a partition that can hold several gigabytes of data. That is, if you use the default directory ${HOME}/miniconda3/, then installed software counts against your home allocation on Greatlakes. For me this has only been an issue if I'm installing Cuda versions.

Create a conda environment and install ncl

• Create a conda environment using the command conda create --name ${NCL_ENV_NAME} -c conda-forge ncl (or whatever you want your environment to be named-if you forget, you can run conda info --env to find all the conda environments on your system). This only needs to be run once.
• Run conda activate ${NCL_ENV_NAME}. After installation, this must be run any time you want to access the ncl command.

And you're done!

Installing from source:

On a non-x86_64 machine, binaries may not be available from conda. It appears that even the NCL team has not tried building for machines with an ARM-based M1 Mac. However, it appears that the MacPorts package manager may install it from source. This is probably the option that requires the least effort on your part, according to this NCL Github issue.