Reproduce “ELK: Exploring the Efficiency of Inter-core Connected AI Chips with Deep Learning Compiler Techniques” 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
Yiqi Liu, Yuqi Xue, Noelle Crawford, Jilong Xue, Jian Huang
Elk is a compiler that jointly schedules HBM preloads, inter-core communication, and SRAM allocation for inference on inter-core connected AI chips. It selects preload order, memory plans, and timing to balance operator execution with data movement. In the authors’ emulated LLM-decoding evaluation, Elk-Full averaged 94.84% of Ideal performance; Ideal is theoretical, and HBM behavior was emulated.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.