The MIT Programming Languages group has selected “Optimizing Datalog for the GPU” Sun, Y., Shovon, A.R, Gilray, T., Kumar, S., Micinski, K. to be presented and published at the 2026 MIT PLR Workshop.

The MIT Programming Languages Review launched in 2023, to highlight recent research deemed interesting and important, highlighting recent developments with significant potential to shape the future direction of PL research and/or industry practice. The MIT PL Review program committee consists of senior MIT PhD students, with the mission of selecting papers that can substantially transform the PL community and beyond, with a focus on emerging trends rather than established lines of research.

Abstract:
“Optimizing Datalog for the GPU” explores how modern Datalog engines can be redesigned for accelerator hardware. It introduces a GPU-backed Datalog engine built around a novel data structure called hash-indexed sorted array (HISA). HISA enables efficient execution of sparse relational workloads by combining the benefits of constant-time indexed access with dense, range-structured data layouts. By rethinking how a foundational declarative programming technology maps to GPUs, the work highlights a growing intersection between programming languages, program analysis infrastructure, and accelerator-aware system design, illustrating how PL implementation techniques can evolve alongside modern hardware trends.

The Workshop will be held May 8, 2026.