Performance Benchmarks and Optimizations of a Gravitational-Wave Simulation Software

Project Information

astrophysics, c++, matlab
Project Status: Complete
Project Region: CAREERS
Submitted By: Steven Liebling
Project Email: liebling@gmail.com
Project Institution: LIU Post
Anchor Institution: CR-Rensselaer Polytechnic Institute

Students: MANAS VISHAL

Project Description

Next-generation gravitational-wave (GW) detectors, such as the Laser Interferometer Space Antenna (LISA), will be able to detect GW signals from extreme mass ratio inspirals. Numerical simulation of these systems requires extremely long-time integration (~100,000 orbital cycles) and high accuracy for achieving the full scientific potential of LISA. We are developing a production-level simulation code for this purpose. The code is currently in Matlab, and initial work to rewrite it in C++17 is underway. This project aims to build a modernized software development ecosystem for the C++17 code that includes performance benchmarking tests and optimizations. In this project, the student will write benchmark tests that time computationally intensive parts of the code. The tests will be managed by the ctest tool that ships with cmake. We plan to run our performance tests on different HPC machines, including the MGHPCC and our local machine Carnie. The student will also explore single-core and single-node performance optimizations whose effectiveness will be assessed according to the benchmark tests.

Project Information

astrophysics, c++, matlab
Project Status: Complete
Project Region: CAREERS
Submitted By: Steven Liebling
Project Email: liebling@gmail.com
Project Institution: LIU Post
Anchor Institution: CR-Rensselaer Polytechnic Institute

Students: MANAS VISHAL

Project Description

Next-generation gravitational-wave (GW) detectors, such as the Laser Interferometer Space Antenna (LISA), will be able to detect GW signals from extreme mass ratio inspirals. Numerical simulation of these systems requires extremely long-time integration (~100,000 orbital cycles) and high accuracy for achieving the full scientific potential of LISA. We are developing a production-level simulation code for this purpose. The code is currently in Matlab, and initial work to rewrite it in C++17 is underway. This project aims to build a modernized software development ecosystem for the C++17 code that includes performance benchmarking tests and optimizations. In this project, the student will write benchmark tests that time computationally intensive parts of the code. The tests will be managed by the ctest tool that ships with cmake. We plan to run our performance tests on different HPC machines, including the MGHPCC and our local machine Carnie. The student will also explore single-core and single-node performance optimizations whose effectiveness will be assessed according to the benchmark tests.