Reproduce “Retargetable Compilation for Tensor Kernels via a Unified Layout Abstraction” 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
Likai Ma, Renze Chen, Tianqi Li, Yuhao Luo, Youwei Xiao, Size Zheng, Xiuhong Li, Yun Liang
Arknife makes instruction operand layouts explicit, then checks requirements across a program, inserts and removes repairs, and solves compatible mappings for target instructions. In one Hexagon quantized-add example, it removes three provisional permutations. Across nine x86 and nine Hexagon workloads, it reports geometric-mean speedups over MISAAL of 5.08× and 4.53×; these are suite results, not guarantees for each workload.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.