Reproduce “SPRING: Systematic Profiling of Randomly Interconnected Neural Networks Generated by HLS” 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
Rui Shi, Seda Ogrenci
Northwestern University researchers develop SPRING, a profiling framework for FPGA-based streaming machine learning accelerators that enables concurrent monitoring of over 200 internal signals while maintaining synchronization with data streams, integrated into the hls4ml open-source tool for automated analysis of randomly interconnected neural networks.
Key insights
- FIFO size correlates with layer types and code structure
- Small RINNs sufficient to reveal most FIFO size requirements
- Resource overhead can be managed by replacing long profiling streams with dummy ones
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.