Search papers, labs, and topics across Lattice.
This paper investigates the applicability of roofline modeling for sparse matrix-dense matrix multiplication (SpMM), a key kernel in various domains, by explicitly considering the impact of matrix sparsity structure. They evaluate CSR, CSB, and MKL implementations on matrices from the SuiteSparse collection, grouped by sparsity pattern. The authors derive sparsity-aware roofline models that account for memory traffic, cache locality, and blocking, demonstrating that a single model fails to accurately predict performance across diverse structures.
A single roofline model is insufficient for predicting SpMM performance across diverse sparsity structures; data layout and blocking strategies must be evaluated in the context of matrix structure.
Sparse matrix-dense matrix multiplication (SpMM) is a critical kernel in scientific computing, graph analytics, and machine learning, whose performance is often constrained by memory bandwidth. In this work, we investigate the applicability and limitations of roofline modeling for SpMM by explicitly accounting for the impact of matrix sparsity structure on arithmetic intensity and attainable performance. We evaluate three SpMM implementations: Compressed Sparse Row (CSR), Compressed Sparse Blocks (CSB), and Intel's Math Kernel Library (MKL). Each implementation was tested using large-scale matrices from the SuiteSparse collection and grouped by sparsity pattern, including block-structured, banded (diagonal), scale-free, and uniformly random matrices. We derive sparsity-aware roofline models that incorporate memory traffic, cache locality, and blocking behavior, and demonstrate that a single model is insufficient to accurately predict performance across diverse structures. Experiments were conducted on an AMD-based Perlmutter compute node with a varying number of columns in the dense matrix. In particular, blocking and structured sparsity significantly alter effective arithmetic intensity. The results show that accurate roofline-based performance analysis of SpMM requires sparsity-aware modeling, and that data layout and blocking strategies must be evaluated in the context of matrix structure rather than through a single unified model.
