The Chunks and Tasks Matrix Library 2.0

TL;DR

This paper introduces a C++ library for parallel sparse matrix computations using a novel data structure and programming model, enabling efficient distributed processing and dynamic load balancing across various matrix sparsity patterns.

Contribution

The library combines sparse quadtree representation with the Chunks and Tasks model to improve parallel sparse matrix algorithms and dynamic load balancing.

Findings

Successfully demonstrates dynamic load balancing across different matrix structures

Achieves efficient distributed memory parallelization of sparse matrix algorithms

Utilizes the CHT-MPI 2.0 runtime for dynamic data locality exploitation

Abstract

We present a C++ header-only parallel sparse matrix library, based on sparse quadtree representation of matrices using the Chunks and Tasks programming model. The library implements a number of sparse matrix algorithms for distributed memory parallelization that are able to dynamically exploit data locality to avoid movement of data. This is demonstrated for the example of block-sparse matrix-matrix multiplication applied to three sequences of matrices with different nonzero structure, using the CHT-MPI 2.0 runtime library implementation of the Chunks and Tasks model. The runtime library succeeds to dynamically load balance the calculation regardless of the sparsity structure.