Thomas Koehler

I am a PhD student in the School of Computing Science at the University of Glasgow in Scotland, supervised by Michel Steuwer and Phil Trinder.

Before coming to Glasgow I was a student at Sorbonne Université in Paris, France.

You can read my Curriculum Vitae to learn more.

A Domain-Extensible Compiler that Optimizes for Evolving Algorithms and Hardware

Recent breakthroughs in machine intelligence have sparked ambitious visions for the future where intelligent machines are ubiquitous: they diagnose diseases, drive vehicles, schedule factories, predict climate change, and more. To realize this future, massive computing power is required, but physical resources are limited. In the cloud, huge data centres are taking an increasing toll on our planet. At the edge, running advanced neural networks on small devices with low energy supply is not yet possible.

If we want to keep up the pace of progress in a sustainable way, we must use resources more efficiently through continuous algorithm, hardware and software optimization. These three optimization axes are inter-dependent: in particular, each new algorithm and hardware architecture requires new software optimizations.

My research is about exploring the design of compilers that can adapt to new algorithms and hardware while generating high-performance code: domain-extensible compilers that provide an extensible set of program abstractions and optimizations. More specifically, I am the lead developer of the Rise project, which combines a high-level functional language with a system of rewrite rules to encode optimization choices.