Ocean : Fast Estimation-Based Sparse General Matrix-Matrix Multiplication on GPU Conference: 2026 International Conference on Supercomputing; July 06–09, 2026; Belfast, United Kingdom 2026 International Conference on Supercomputing (ICS ’26), July 06–09, 2026, Belfast, United Kingdom DOI: 10.1145/3797905.3807868 ISBN: 979-8-4007-2522-7/2026/07 CCS: Theory of computation Massively parallel algorithms CCS: Computing methodologies Linear algebra algorithms
Yifan Li, Giulia Guidi · 2026
Read Ocean : Fast Estimation-Based Sparse General Matrix-Matrix Multiplication on GPU Conference: 2026 International Conference on Supercomputing; July 06–09, 2026; Belfast, United Kingdom 2026 International Conference on Supercomputing (ICS ’26), July 06–09, 2026, Belfast, United Kingdom DOI: 10.1145/3797905.3807868 ISBN: 979-8-4007-2522-7/2026/07 CCS: Theory of computation Massively parallel algorithms CCS: Computing methodologies Linear algebra algorithms by Yifan Li, Giulia Guidi (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-19004.txt.
Description
Li and Guidi present Ocean, an estimation based sparse matrix multiplication workflow for GPUs. Sparse matrix multiplication drives simulations, graph analytics, and machine learning, but irregular access and imbalance hurt GPU speed. Current tools use two passes, with the symbolic pass alone taking roughly 28 percent of runtime. Ocean replaces exact symbolic computation with lightweight HyperLogLog estimators, a strategy that picks the best workflow, and a hybrid accumulator. It beats leading GPU implementations with speedups up to 2.8 times on A100 and H100 hardware.
Catalog record
| CREATOR | Yifan Li, Giulia Guidi |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 67857 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.19004