Reproduce “LightCode: Compiling LLM Inference for Photonic-Electronic Systems” 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

Ryan Tomich, Zhizhen Zhong, Dirk Englund

LightCode maps tensor operations in LLM prefill across electronic GPUs and photonic accelerators, accounting for computation and data-transfer costs to optimize latency or energy. Simulations show energy-optimized mappings often offload linear operations to photonic hardware, while latency-optimized mappings often retain them on GPUs. The abstract reports up to 50% lower simulated energy, without identifying the model or baseline.

Summary by alphaXiv.

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