MPI4py-based Parallel Implementation of the Richardson-Lucy Algorithm

Documentation and code repository for PHYS 244: Parallel Computing

Directory Contents

• assets: hosts auxiliary files such as the image files of the results shown below.
• code: contains python scripts for a point-source toy model, $^{44}$Ti (point-source) and positron (diffuse-source) imaging analyses.
• docs: references used to write the code.

Dependencies

• numpy
• mpi4py
• h5py
• cosipy and histpy (optional)

Result

The code successfully recovers the injected signal of 3 supernovae, Cassiopeia A, G1.9+0.3 (Galactic Center), and SN1987A (Large Magellanic Cloud) in just 12 seconds. This is a 5x speed-up with respect to the code in the COSIpy library, which took ~1 minute. The parallel implementation also circumvents the need to load extremely large onto memory.