Helping to make LuaJIT faster

Laurence Tratt, a programmer and Reader in Software Development at King's College London, discusses the complexity of advanced VMs with Just-In-Time (JIT) compilers. These VMs observe a program's run-time behavior to compile frequently executed parts into machine code, leading to significant speed-ups. However, they are internally complex and can have deoptimization bugs. Tratt's research group at King's College London has received funding from Cloudflare to improve LuaJIT's performance. The team will leverage their expertise in VM work and benchmarking to test hypotheses about heuristic interactions and the importance of rigorous pre- and post-change benchmarking. They aim not only to improve LuaJIT but also contribute to broader research on understanding how VM performance can be improved, potentially influencing future generations of VMs.