Affiliation:
1. University of Illinois at Urbana--Champaign, Urbana, IL
2. Google, Mountain View, CA
Abstract
Sparse matrix--matrix multiplication (SpGEMM) is a key operation in numerous areas from information to the physical sciences. Implementing SpGEMM efficiently on throughput-oriented processors, such as the graphics processing unit (GPU), requires the programmer to expose substantial fine-grained parallelism while conserving the limited off-chip memory bandwidth. Balancing these concerns, we decompose the SpGEMM operation into three highly parallel phases: expansion, sorting, and contraction, and introduce a set of complementary bandwidth-saving performance optimizations. Our implementation is fully general and our optimization strategy adaptively processes the SpGEMM workload row-wise to substantially improve performance by decreasing the work complexity and utilizing the memory hierarchy more effectively.
Publisher
Association for Computing Machinery (ACM)
Subject
Applied Mathematics,Software
Cited by
68 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献
1. A Survey of Accelerating Parallel Sparse Linear Algebra;ACM Computing Surveys;2023-08-28
2. Efficient Execution of SpGEMM on Long Vector Architectures;Proceedings of the 32nd International Symposium on High-Performance Parallel and Distributed Computing;2023-08-07
3. HASpGEMM: Heterogeneity-Aware Sparse General Matrix-Matrix Multiplication on Modern Asymmetric Multicore Processors;Proceedings of the 52nd International Conference on Parallel Processing;2023-08-07
4. The Resource Problem of Using Linear Layer Leakage Attack in Federated Learning;2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR);2023-06
5. A Machine Learning Approach Towards Runtime Optimisation of Matrix Multiplication;2023 IEEE International Parallel and Distributed Processing Symposium (IPDPS);2023-05