Reproduce “Taming Multi-Dimensional Skew in Sparse Matrix Multiplication with Contention-Aware Scheduling [Technical Report]” 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
Wentao Huang, Qiming Huang, Yunhong Ji, Mian Lu, Yuqiang Chen, Bingsheng He, Kian-Lee Tan
The AMS framework schedules consecutive-row morsels sized by expected output, then orders them to balance work while limiting cache contention. On a highly skewed G500 R-MAT matrix-square workload at scale 19 and edgefactor 2, AMS-HASH achieves a 1.44× throughput speedup over HASH. The authors report comparable performance on non-skewed workloads, while morsel-sorting overhead can hurt very sparse cases.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.