Reproduce “Mitigating inefficient task mappings with an Adaptive Resource-Moldable Scheduler (ARMS)” 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
Mustafa Abduljabbar, Mahmoud Eljammaly, Miquel Pericas
ARMS schedules tasks by learning how task functions perform at logical data locations, then selecting both placement and thread width. Its online history-based model uses task type and a topology key, or DAG position when data location is unavailable. In a synthetic MatMul chain, ARMS-M was reported about 3.5× faster than ADWS at DAG parallelism 2–8; the advantage narrowed near 32-worker parallelism.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.