Open Call project, 2023 rubix
GPU computing in Astronomy
Total developer time: 2 months
Contact person: Dr. Tobias Buck
Outline
The goal of this project is to develop a differentiable forward model for creating mock observations in astronomy. Similar systems already exist, e.g. SimSpin in R, but are not able to leverage modern GPU accelerator hardware or seamlessly interface with machine learning frameworks and tools. This project thus aims at developing a modern and extendible system for creating astronomical mock observations in JAX that massively reduces the runtime of currently available methods for creating mock observations which then opens the way for adding capabilities that until now have been unavailable.
Role of the SSC
- Prototyping: Propose and implement a suitable software architecture compatible with
JAX's paradigms - RSE Education: Introduce involved scientists to best practices in software development.
- Support: Help with implementing modern development workflows in their team
Results
- Successful implementation of sustainable development workflows for the researchers
- Successful implementation of software development best pracitces in the research group
- Increase in performance over state-of-the-art codes by a factor of ~600