Getting parallel bilby production-ready for LIGO’s fourth observing run

CI: – Dr. A. Vajpeyi

Parallel Bilby (pBilby) is a highly parallelized gravitational-wave (GW) inference code that can reduce the wall time of inference almost linearly with the number of parallel processes. pBilby has been used to make inferences on several “exceptional” GW events from LIGO’s third observing run (e.g., the first intermediate-mass black hole event, GW190521). The inferences pBilby can make are currently impossible with other existing codes, and hence this software has already been used in more than 30 publications. However, due to the poor code quality and lack of a testing framework, maintenance and further development of pBilby has become challenging. To allow pBilby to remain a state-of-art software, pBilby’s development team needs better development practices.
Project Goals:
We would like ADACS to:
Establish good development practices within the (currently small) development team to make the development accessible to new developers.
Demonstrate how the development team can add tests and refactor the codebase using software engineering principles
With better development practices and workflows:
New contributors may easily be inducted into the development workflow while keeping the code-base manageable.
Optimizations to the software to accelerate inference may be possible without fear of introducing new errors (researchers have various ideas that may boost the speed of inference by several factors. See [1,2])
In summary, these outcomes will enable researchers to improve pBilby to analyze GW data faster and with higher precision, a necessity as the number of GW events keeps increasing.