A Software-Based Fault Tolerance Mechanism for Matrix Multiplication Operations in Tensor Cores
IEEE Transactions on Computers · 2026-09 · 75 (9) · 3212-3225
DOI: 10.1109/tc.2026.3702572 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Computers, 2026-09. Verfasst von Robert Limas Sierra und 3 weiteren Autor:innen. Die Arbeit nennt 45 Literaturangaben. Gefördert von: National Resilience and Recovery Plan, National Center for HPC, Big Data and Quantum Computing.
Abstract (Originalsprache)
Modern Graphics Processing Units (GPUs) are increasingly employed to enhance the performance of algorithms across scientific and machine learning domains. Given the importance of General Matrix Multiplication (GEMM) operations, GPUs feature specialized in-chip accelerators, such as Tensor Cores (TCUs), to speed them up. High-Performance Computing (HPC) and safety-critical sectors (e.g., automotive, space, and autonomous robotics) impose severe constraints concerning not only energy consumption, performance, and area but also reliability. Faults arising from advanced semiconductor technologies or sustained HPC workloads can silently propagate, potentially leading to catastrophic failures. This work introduces a hardware-aware, software-based fault tolerance method to enhance the resilience of GEMM operations on TCUs. By leveraging TCU architecture and parallel operation distribution, the method enables efficient fault detection and mitigation. It utilizes redundant executions on TCU arithmetic cores (Dot-Product Units) to detect and, if required, correct fault effects. The method’s flexibility supports online detection and correction of transient and permanent hardware faults in TCU’s arithmetic units. Experimental results on real GPUs show that the proposed mechanism introduces minimal and constant memory overhead and a negligible performance overhead (up to 1.13 times) across operand sizes. Thus, this solution offers an effective and complementary hardening strategy for TCU operations.
Förderung
- National Resilience and Recovery Plan
- National Center for HPC
- Big Data and Quantum Computing
Zitieren
APA
Sierra, R. L., Guerrero-Balaguera, J., Condia, J. E. R., & Reorda, M. S. (2026). A Software-Based Fault Tolerance Mechanism for Matrix Multiplication Operations in Tensor Cores. IEEE Transactions on Computers, 75(9), 3212-3225. https://doi.org/10.1109/tc.2026.3702572
Vancouver
Sierra RL, Guerrero-Balaguera J, Condia JER, Reorda MS. A Software-Based Fault Tolerance Mechanism for Matrix Multiplication Operations in Tensor Cores. IEEE Transactions on Computers. 2026;75(9):3212-3225. doi:10.1109/tc.2026.3702572
BibTeX
@article{Sierra2026,
author = {Sierra, Robert Limas and Guerrero-Balaguera, Juan-David and Condia, Josie E. Rodriguez and Reorda, Matteo Sonza},
title = {A Software-Based Fault Tolerance Mechanism for Matrix Multiplication Operations in Tensor Cores},
journal = {IEEE Transactions on Computers},
year = {2026},
volume = {75},
number = {9},
pages = {3212--3225},
doi = {10.1109/tc.2026.3702572}
}
EUVIMED