SpComm3D: A Framework for Enabling Sparse Communication in 3D Sparse Kernels

TL;DR

SpComm3D is a new framework that enables efficient, sparsity-aware communication in 3D sparse kernels, significantly reducing communication, memory usage, and runtime for distributed ML computations.

Contribution

It introduces a sparsity-aware communication framework for 3D sparse kernels, improving scalability and efficiency over existing sparsity-agnostic methods.

Findings

Up to 20x reduction in communication, memory, and runtime.

Superior scalability demonstrated on up to 1800 processors.

Effective for key ML kernels: SDDMM and SpMM.

Abstract

Existing 3D algorithms for distributed-memory sparse kernels suffer from limited scalability due to reliance on bulk sparsity-agnostic communication. While easier to use, sparsity-agnostic communication leads to unnecessary bandwidth and memory consumption. We present SpComm3D, a framework for enabling sparsity-aware communication and minimal memory footprint such that no unnecessary data is communicated or stored in memory. SpComm3D performs sparse communication efficiently with minimal or no communication buffers to further reduce memory consumption. SpComm3D detaches the local computation at each processor from the communication, allowing flexibility in choosing the best accelerated version for computation. We build 3D algorithms with SpComm3D for the two important sparse ML kernels: Sampled Dense-Dense Matrix Multiplication (SDDMM) and Sparse matrix-matrix multiplication (SpMM). Experimental evaluations on up to 1800 processors demonstrate that SpComm3D has superior scalability and outperforms state-of-the-art sparsity-agnostic methods with up to 20x improvement in terms of communication, memory, and runtime of SDDMM and SpMM. The code is available at: https://github.com/nfabubaker/SpComm3D