English

Multilayer Dataflow: Orchestrate Butterfly Sparsity to Accelerate Attention Computation

Hardware Architecture 2024-11-26 v2

Abstract

Recent neural networks (NNs) with self-attention exhibit competitiveness across different AI domains, but the essential attention mechanism brings massive computation and memory demands. To this end, various sparsity patterns are introduced to reduce the quadratic computation complexity, among which the structured butterfly sparsity has been proven efficient in computation reduction while maintaining model accuracy. However, its complicated data accessing pattern brings utilization degradation and makes parallelism hard to exploit in general block-oriented architecture like GPU. Since the reconfigurable dataflow architecture is known to have better data reusability and architectural flexibility in general NN-based acceleration, we want to apply it to the butterfly sparsity for acquiring better computational efficiency for attention workloads. We first propose a hybrid butterfly-sparsity network to obtain better trade-offs between attention accuracy and performance. Next, we propose a scalable multilayer dataflow method supported by coarse-grained streaming parallelism designs, to orchestrate the butterfly sparsity computation on the dataflow array. The experiments show that compared with Jetson Xavier NX, our design has a speedup of up to 14.34×14.34\times (9.29×9.29\times on average) as well as 11.14×11.14\times energy efficiency advancement in attention workloads. In comparison with SOTA attention accelerators of the same peak performance, our dataflow architecture acquires 2.38×2.38\times-4.7×4.7\times efficiency improvement as well as 6.60×6.60\times-15.37×15.37\times energy reduction with butterfly sparsity optimization.

Keywords

Cite

@article{arxiv.2411.00734,
  title  = {Multilayer Dataflow: Orchestrate Butterfly Sparsity to Accelerate Attention Computation},
  author = {Haibin Wu and Wenming Li and Kai Yan and Zhihua Fan and Peiyang Wu and Yuqun Liu and Yanhuan Liu and Ziqing Qiang and Meng Wu and Kunming Liu and Xiaochun Ye and Dongrui Fan},
  journal= {arXiv preprint arXiv:2411.00734},
  year   = {2024}
}

Comments

9 pages, 17 figures, ISCA 2025, 2024/11/23, Butterfly Sparsity Optimization Using Dataflow

R2 v1 2026-06-28T19:44:31.192Z