Stefan Marr

Associate Professor (Senior Lecturer)
and Royal Society Industry Fellow

Stefan Marr I am working at the School of Computing at the University of Kent. My work focuses on programming language implementation techniques, complex concurrent systems, and tooling.

Much of work revolves around making interpreters faster, detecting and preventing concurrency bugs at run time, and use modern language runtime systems to provide programmers with better tooling.

For contact details see my page at the School of Computing.

Archive of posts with the tag Inline Cache

Nov 8, 2022: How Effective are Classic Lookup Optimizations for Rails Apps?

We know that Ruby and especially Rails applications can be very dynamic and pretty large. Though, many of the optimizations interpreters and even just-in-time compilers use have been invented in the 1980s and 1990s before Ruby and Rails even existed. So, I was wondering: do these optimizations still have a chance of coping with the millions of lines of Ruby code that large Rails apps from Shopify, Stripe, or GitLab have? Unfortunately, we don’t have access to such applications. As the next best thing, we took the largest Ruby benchmarks we could get our hands on, and analyzed those.
Oct 30, 2022: Reducing Memory Footprint by Minimizing Hidden Class Graphs

Tomoharu noticed in his work on the eJSVM, a JavaScript virtual machine for embedded systems, that quite a bit of memory is needed for the data that helps us to represent JavaScript objects efficiently. So, we started to look into how the memory use could be reduced without sacrificing performance.
Apr 28, 2015: Zero-Overhead Metaprogramming

Runtime metaprogramming and reflection are slow. That’s a common wisdom. Unfortunately. Using refection for instance with Java’s reflection API, its dynamic proxies, Ruby’s #send or #method_missing, PHP’s magic methods such as __call, Python’s __getattr__, C#’s DynamicObjects, or really any metaprogramming abstraction in modern languages unfortunately comes at a price. The fewest language implementations optimize these operations. For instance, on Java’s HotSpot VM, reflective method invocation and dynamic proxies have an overhead of 6-7x compared to direct operations.