POPL 2026
Sun 11 - Sat 17 January 2026 Rennes, France
Sun 11 Jan 2026 09:53 - 10:03 at Salle 13 - First Session Chair(s): Alexander K. Lew

We present a design for a probabilistic programming system that enables massively parallel Gibbs sampling on GPUs through workload-aware scheduling of hierarchical parallelism. Workload characteristics of the Gibbs sampler dictate where parallelism is most effective: across variables, within variable updates, or hierarchically on both levels. Hence, there are no one-size-fits-all solutions for efficient parallel Gibbs sampling. Our system addresses this challenge through two key components: a factor-based intermediate representation that enables static analysis to detect parallelism over variables via graph coloring, and reified schedule objects that allow both manual control and automatic performance tuning of hierarchical parallelism. Preliminary results demonstrate orders of magnitude performance improvements for popular Bayesian networks, Ising models, and hidden Markov models. By combining static analysis with dynamic auto-tuning, our design significantly reduces the development cost of efficient GPU-accelerated Gibbs samplers while maintaining high-level abstractions for model specification.

Sun 11 Jan

Displayed time zone: Brussels, Copenhagen, Madrid, Paris change

09:00 - 10:30
First SessionLAFI at Salle 13
Chair(s): Alexander K. Lew Yale University
09:00
5m
Day opening
Welcome
LAFI
Hugo Paquet Inria, École Normale Supérieure, Alexander K. Lew Yale University
09:07
20m
Industry talk
Basis — A Programming Languages Take on Principled Foundations for AI
LAFI
09:29
10m
Talk
Towards an Equational Calculus of Interventions
LAFI
Shubh Agrawal Northeastern University, Jialu Bao Northeastern University, Steven Holtzen Northeastern University
Pre-print
09:41
10m
Talk
Typed Abstractions for Causal Probabilistic Programming
LAFI
Theo Wang University of Cambridge, University of Oxford, Dario Stein University of Oxford, Eli Bingham Broad Institute, Jack Feser Basis, Ohad Kammar University of Edinburgh, Michael Lee University of Cambridge, UK, Jeremy Yallop University of Cambridge
File Attached
09:53
10m
Talk
A Design for Massively Parallel Gibbs Sampling on the GPU via Static and Dynamic Analysis of Probabilistic Programs
LAFI
Matin Ghavami Massachusetts Institute of Technology, Martin C. Rinard Massachusetts Institute of Technology, Vikash Mansinghka Massachusetts Institute of Technology
10:05
10m
Talk
A Design Proposal for GraPPL: Probabilistic Programming with Low-Level, High-Performance GPU Programmable Inference
LAFI
Karen Chung Massachusetts Institute of Technology, Elias Rojas Collins MIT, McCoy Reynolds Becker MIT, Mathieu Huot MIT, Vikash Mansinghka Massachusetts Institute of Technology