Parallelisation of the SoFiA Source Finding Pipeline

CI: – Dr. Tobias Westmeier

The aim of this proposal is to get support for parallelising the Source Finding Application (SoFiA), a new source finding and parameterisation pipeline developed for the automated detection of galaxies in some of the upcoming SKA precursor and pathfinder spectral-line surveys, most notably the WALLABY all-sky HI survey on ASKAP. In preparation for this task, the entire pipeline has already been converted from Python/C++ to plain C by the Lead CI. The scope of this project is to employ tools such as MPI or OpenMP to implement multi-threading of the most time-consuming algorithms that operate on the data arrays to speed up processing by utilising all available cores on a single computing node. Examples of time-consuming, yet easily parallelisable algorithms include convolution of the data with Gaussian and Boxcar smoothing kernels or the application of multivariate kernel density estimation to measure the reliability of source detections.
The proposed project will be of great importance to making the SoFiA pipeline capable of processing the large data volumes expected to be produced by ASKAP survey projects such as WALLABY. As SoFiA has been made publicly available, the proposed parallelisation of SoFiA will benefit the entire radio astronomy community in Australia who be able to utilise the software for their own source finding needs. SoFiA has already become the de-facto standard source finding pipeline for radio astronomical spectral-line data, and parallelisation of SoFiA would therefore appear to be an excellent investment of resources.