Reproduce “Predictable Accelerator Design with Time-Sensitive Affine Types” 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

Rachit Nigam, Sachille Atapattu, Samuel Thomas, Zhijing Li, Theodore Bauer, Yuwei Ye, Apurva Koti, Adrian Sampson and 1 more

Dahlia makes FPGA accelerator resource use explicit, using types to reject memory-access conflicts before code reaches a high-level synthesis compiler. Programmers express banking and logical time steps rather than relying on hidden sharing decisions. In a blocked matrix-multiplication exploration, it accepted 354 of 32,000 configurations, primarily retaining Pareto trade-offs, but does not model total FPGA resources.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.