About

I am John LAM (林俊傑, Lam Chun Kit), a PhD student of TACO Lab and ALPACAS group of HKUST under the supervision of Prof. Parreaux and Prof. Goharshady. My research interest is mainly about type system and static analysis for program optimization. I also use techniques from parameterized algorithm for speeding up program analysis, to make the analysis applicable to large scale real world projects.

Abstraction allows us to build more complex programs while maintaining sanity, but this can also hide performance issues, such as accidentally implementing a linear time algorithm with quadratic time complexity, or using the wrong memory causing a 100x slowdown. I believe that with better language design and program optimization, we can make it easier to create correct and efficient programs.

However, static analysis required for program optimization are often computationally expensive, making it hard to scale to programs with over thousand lines of code. In fact, basic problems such as register allocation and instruction scheduling are NP Hard. Existing tools often use heuristics to avoid this issue, which are sometimes brittle and far from optimal. With parameterized algorithms, we can exploit the inherent properties of these problems, such as the control flow graph of C goto-free programs have bounded treewidth, to devise better algorithms that are feasible for larger instances.

Publications

I'm interested in developing algorithms for static analysis, language design, as well as software engineering related to performance optimization in general.

Authors are listed in alphabetical order.

1. Fast and Optimal Extraction for Sparse Equality Graphs
   Amir Kafshdar Goharshady, Chun Kit Lam, Lionel Parreaux
   OOPSLA 2024
2. Practical Constant-Time Memory Management for Functional Programming
   Chun Kit Lam, Lionel Parreaux
   FLOPS 2024 (preprint, paper)
3. The Bounded Pathwidth of Control-Flow Graphs
   Giovanna Kobus Conrado, Amir Kafshdar Goharshady, Chun Kit Lam
   OOPSLA 2023 (paper)

Unrelated Works

I also work on the manifold library for fast mesh boolean, as well as electronics and mechanical DIY projects.