GUST: Graph Edge-Coloring Utilization for Accelerating Sparse Matrix Vector Multiplication

TL;DR

GUST is a hardware/software co-design that improves sparse matrix-vector multiplication performance by separating multipliers and adders, enabling resource sharing and using a novel edge-coloring scheduling algorithm to reduce collisions.

Contribution

GUST introduces a novel hardware/software co-design with separated hardware units and an edge-coloring scheduling algorithm to enhance hardware utilization in SpMV.

Findings

Achieves an average hardware utilization of 33.67%.

Effectively reduces resource collisions in sparse matrix processing.

Outperforms prior domain-specific architectures on real-world datasets.

Abstract

Sparse matrix-vector multiplication (SpMV) plays a vital role in various scientific and engineering fields, from scientific computing to machine learning. Traditional general-purpose processors often fall short of their peak performance with sparse data, leading to the development of domain-specific architectures to enhance SpMV. Yet, these specialized approaches, whether tailored explicitly for SpMV or adapted from matrix-matrix multiplication accelerators, still face challenges in fully utilizing hardware resources as a result of sparsity. To tackle this problem, we introduce GUST, a hardware/software co-design, the key insight of which lies in separating multipliers and adders in the hardware, thereby enabling resource sharing across multiple rows and columns, leading to efficient hardware utilization and ameliorating negative performance impacts from sparsity. Resource sharing, however, can lead to collisions, a problem we address through a specially devised edge-coloring scheduling algorithm. Our comparisons with various prior domain specific architectures using real-world datasets shows the effectiveness of GUST, with an average hardware utilization of $33.67\%$.