Optimising Julia code on GPUs: Case studies from ocean and atmosphere modelling

Speaker: Mikolaj Kowalski, Jack Franklin, David Kamm - RSEs, ICCS, University of Cambridge
Series: RSE Seminars
Also open online: the joining details are on the talk's page.

Julia is a just-in-time (JIT) compiled language with a strong focus on scientific computing and performance. It has gained significant popularity in the climate science community, largely due to its flexibility and excellent native support for GPU computing. Over the last year, our team at the Institute for Computing for Climate Science (ICCS) has collaborated closely with the developers of major Julia-based ocean and atmosphere models to optimize their performance on GPU architectures.

This talk presents a series of case studies derived from these collaborations, identifying the specific speed-up opportunities we discovered and exploited. We will dive into the technical details of how Julia compiles and executes code on GPUs, highlighting where performance optimization aligns with traditional patterns in languages like C++ or Fortran, and where the JIT-based Julia ecosystem requires a distinct approach.

We conclude by showcasing the distinct impact of the RSE mindset on research software. By bringing a systematic approach and specialized computing expertise to the collaboration, we are able to spot subtle performance anomalies that lead to transformative speed-ups. This talk illustrates how the RSE's focus on engineering best practices and hardware-specific details directly enhances the capability and scale of modern scientific discovery.

Zoom link