New Row-grouped CSR format for storing the sparse matrices on GPU with implementation in CUDA

TL;DR

This paper introduces a new GPU-optimized sparse matrix storage format called Row-grouped CSR, implemented in CUDA, and compares its performance against the Hybrid format across 1,600 matrices.

Contribution

The paper presents a novel sparse matrix format tailored for GPUs, with detailed implementation and performance comparison against existing formats.

Findings

The new format performs better on certain matrix types.

It shows advantages in both strong and weak scaling scenarios.

Performance varies depending on matrix characteristics.

Abstract

In this article we present a new format for storing sparse matrices. The format is designed to perform well mainly on the GPU devices. We present its implementation in CUDA. The performance has been tested on 1,600 different types of matrices and we compare our format with the Hybrid format. We give detailed comparison of both formats and show their strong and weak parts.