FLAASH: Flexible Accelerator Architecture for Sparse High-Order Tensor Contraction

TL;DR

FLAASH is a flexible hardware accelerator designed for efficient sparse high-order tensor contractions in machine learning, achieving over 25x speedup by distributing computations across specialized engines.

Contribution

The paper introduces a modular accelerator architecture that supports customizable memory and job distribution for sparse tensor contraction, addressing control flow and high-sparsity challenges.

Findings

Achieves over 25x speedup on deep learning workloads

Effectively handles high-order and high-sparsity tensor contractions

Demonstrates significant performance improvements with increased sparsity and tensor order

Abstract

Tensors play a vital role in machine learning (ML) and often exhibit properties best explored while maintaining high-order. Efficiently performing ML computations requires taking advantage of sparsity, but generalized hardware support is challenging. This paper introduces FLAASH, a flexible and modular accelerator design for sparse tensor contraction that achieves over 25x speedup for a deep learning workload. Our architecture performs sparse high-order tensor contraction by distributing sparse dot products, or portions thereof, to numerous Sparse Dot Product Engines (SDPEs). Memory structure and job distribution can be customized, and we demonstrate a simple approach as a proof of concept. We address the challenges associated with control flow to navigate data structures, high-order representation, and high-sparsity handling. The effectiveness of our approach is demonstrated through various evaluations, showcasing significant speedup as sparsity and order increase.