ROME: Maximizing GPU Efficiency for All-Pairs Shortest Path via Taming Fine-Grained Irregularities
Published in Proceedings of the 31st ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming (PPoPP), 2026
ROME targets the all-pairs shortest path (APSP) problem on GPUs and addresses the challenge of fine-grained irregularities in graph workloads. Through novel scheduling strategies and workload reorganization, ROME improves GPU utilization and achieves strong performance on a wide range of graph benchmarks.
Recommended citation: Weile Luo, Yuxin Chen, Xiangrui Yu, Qiang Wang, **Ruibo Fan**, Haibo Liu, et al., "ROME: Maximizing GPU Efficiency for All-Pairs Shortest Path via Taming Fine-Grained Irregularities," in *Proceedings of the 31st ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming (PPoPP)*, 2026.
Download Paper | Code | Download Bibtex