SPIDER: Unleashing Sparse Tensor Cores for Stencil Computation via Strided Swapping

Proceedings of the 31st ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming 2026 Open access

Read full text
Authors: Qian Ping Gu, C.-H. Wu, Heng Shi, Jianguo Yao
DOI: 10.1145/3774934.3786414
Cited by: 1
References: 61

Abstract

Recent research has focused on accelerating stencil computations by exploiting emerging hardware like Tensor Cores. To leverage these accelerators, the stencil operation must be transformed to matrix multiplications. However, this transformation introduces undesired sparsity into the kernel matrix, leading to significant redundant computation.

Fields of Study & Subjects

Computational Physics and Python Applications Geophysical and Geoelectrical Methods Parallel Computing and Optimization Techniques

Publication details

Journal / source
Proceedings of the 31st ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming
Publisher
ACM
Pages
218–231
Publication date
28 January 2026
Language
EN