Periodic Online Testing for Sparse Systolic Tensor Arrays

2025-04-25

AI Summary

Addressing the challenge of detecting permanent faults in sparse systolic tensor arrays for safety-critical machine learning hardware, this paper proposes a lightweight, periodic online self-test method. Executed prior to computation initiation, the method requires only four weight-driven test vectors and leverages preloaded sparse weights to achieve full-array coverage—enabling, for the first time, online self-testing of sparse systolic architectures with zero additional storage overhead. Rigorous evaluation is ensured via gate-level netlist fault injection and structured array modeling. Experimental results across three CNN workloads demonstrate a fault coverage exceeding 99.2%, with computational overhead below 0.3% and negligible area overhead—substantially outperforming existing approaches.

Technology Category

• Machine Learning: Hardware-aware ML
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Natural Language Processing: Safety and Robustness

Application Category

• Security and Privacy: Large-scale security measurements
• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Modern Machine Learning (ML) applications often benefit from structured sparsity, a technique that efficiently reduces model complexity and simplifies handling of sparse data in hardware. Sparse systolic tensor arrays - specifically designed to accelerate these structured-sparse ML models - play a pivotal role in enabling efficient computations. As ML is increasingly integrated into safety-critical systems, it is of paramount importance to ensure the reliability of these systems. This paper introduces an online error-checking technique capable of detecting and locating permanent faults within sparse systolic tensor arrays before computation begins. The new technique relies on merely four test vectors and exploits the weight values already loaded within the systolic array to comprehensively test the system. Fault-injection campaigns within the gate-level netlist, while executing three well-established Convolutional Neural Networks (CNN), validate the efficiency of the proposed approach, which is shown to achieve very high fault coverage, while incurring minimal performance and area overheads.

Problem

• Detecting permanent faults in sparse systolic tensor arrays
• Ensuring reliability of safety-critical ML systems
• Minimizing performance and area overheads during testing

Innovation

• Online error-checking for sparse systolic arrays
• Uses four test vectors for fault detection
• Minimal overhead with high fault coverage