Installation

My recommendation is simply to load the DESI environment:

source /global/common/software/desi/users/adematti/cosmodesi_environment.sh main

Alternatively, you can install pyrecon from here, but there are quite a few dependencies.

pip install git+https://github.com/cosmodesi/pyrecon

Scripts

• analyze_reconstruction.py: Output the r coefficient between true and reconstructed velocity
• correlate_catalog.py: Compute the correlation function or power spectrum
• join_lc_catalog.py: Combine light cone mocks across different redshifts
• prepare_lc_catalog.py: Read light cone mocks and output the necessary fields
• reconstruct_box_catalog.py: Apply reconstruction to a cubic box mock
• reconstruct_lc_catalog.py: Apply reconstruction to a light cone mock
• visualize_correlation.py: Plot the correlation function
• visualize_power.py: Plot the power spectrum

Running on an interactive node

You can start an interactive job on Perlmutter with:

salloc --nodes 1 --qos interactive --time 04:00:00 --constrain cpu

On Cori, cpu needs to be replaced by haswell.

After loading the environment, you can then just run the python command:

python reconstruct_box_catalog.py

When running MPI code (e.g., correlate_catalog.py), you also need to include srun -np N python ..., where N is some reasonable number given the architecture of the cluster (Note that Perlmutter has 128 physical threads, whereas Cori has 32).