Reproduce “Formalizing PTX in Lean to Prove GPU Kernel Correctness” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Luk Koska, Tom Koska, Ben Koska
The authors formalize PTX kernels in Lean by modeling each thread’s execution and checking whether combined events and values satisfy PTX’s memory rules. They prove output contracts for every consistent completed execution under stated launch preconditions, and also show one exists. They report 9 hand-written entry points; real-world assurance assumes no ISA-level undefined behavior and correct compiler and hardware.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.