Reproduce “Hunting CUDA Bugs at Scale with cuFuzz” 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
Mohamed Tarek Ibn ziad, Christos Kozyrakis
cuFuzz fuzzes complete CUDA programs by repeatedly changing inputs to explore new paths and check for errors. It combines host-side and GPU-side coverage, while running incompatible sanitizers in separate processes. Across its evaluation of 14 programs, it reported 43 previously unknown bugs, including 19 in commercial libraries; findings included memory errors, uninitialized reads, and data races.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.