Reproduce “Separation Logic for Memory Conflict Detection in High-Level Synthesis” 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
Yeonseok Lee
A formal framework utilizing Separation Logic at the LLVM IR level is presented to detect memory conflicts from non-affine array indexing in High-Level Synthesis. Developed by SLING AI Inc., this approach includes a Conflict-Free Unrolling condition, SMT-based verification with a conservative fallback, and a soundness theorem to facilitate precise and safe parallel hardware generation.
Key insights
- Separation Logic's property of asserting disjoint memory regions can be adapted to formally model physical memory banks and detect structural conflicts in HLS array partitioning.
- The problem of detecting memory conflicts, even with non-affine arithmetic, can be effectively reduced to algebraic inequality problems solvable by SMT solvers, providing more precise compile-time analysis than affine-limited methods.
- Formal guarantees, such as a Theorem of Soundness, are essential to bridge abstract logical verification with the concrete safety requirements of synthesized Register Transfer Level (RTL) hardware.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.