Reproduce “Scalable Coherent Optical Crossbar Architecture using PCM for AI Acceleration” 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
Daniel Sturm, Sajjad Moazeni
The authors propose a modeled optical accelerator for CNN inference that stores weights in nonvolatile phase-change material within a coherent photonic crossbar. The system model includes memory, optical losses, peripheral electronics, programming time, and chip area. For ResNet50, it estimates 36,382 inferences per second at 30 W and 121 mm², under INT6, batch-size-32 assumptions; no fabricated-chip measurement is reported.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.