Dynasor: A Dynamic Memory Layout for Accelerating Sparse MTTKRP for Tensor Decomposition on Multi-core CPU

TL;DR

This paper presents Dynasor, a novel dynamic memory layout and algorithm for sparse tensor decomposition that significantly accelerates spMTTKRP computations on multi-core CPUs by optimizing data locality, load balancing, and communication overhead.

Contribution

It introduces a new algorithm and tensor format that improve execution speed and efficiency of sparse tensor decomposition on multi-core CPUs.

Findings

Achieves 2.12x to 9.01x speedup over state-of-the-art implementations.

Effectively exploits data locality and load balancing in sparse tensor computations.

Reduces communication overhead through dynamic tensor remapping.

Abstract

Sparse Matricized Tensor Times Khatri-Rao Product (spMTTKRP) is the most time-consuming compute kernel in sparse tensor decomposition. In this paper, we introduce a novel algorithm to minimize the execution time of spMTTKRP across all modes of an input tensor on multi-core CPU platform. The proposed algorithm leverages the FLYCOO tensor format to exploit data locality in external memory accesses. It effectively utilizes computational resources by enabling lock-free concurrent processing of independent partitions of the input tensor. The proposed partitioning ensures load balancing among CPU threads. Our dynamic tensor remapping technique leads to reduced communication overhead along all the modes. On widely used real-world tensors, our work achieves 2.12x - 9.01x speedup in total execution time across all modes compared with the state-of-the-art CPU implementations.