This project includes a selection of toy models for Python.
This toy model has been made to experiment with the numerical integrators in GSL to compute the spontaneous recombination coefficient for macro atoms in the MCRT program Python.
The hope is that we can find a good compromise between accuracy and speed, as the current method in use is becoming too expensive.
This toy model is used to experiment with using remote memory access (RMA)/node sharing/shared memory in MPI to reduce the memory footprint of Python in MPI. The core problem is there are large structures in Python which do not need to be duplicated for each rank. With RMA, only one rank needs to read in, for example, atomic data and all the other ranks can access the data from the root rank.