Neutron radiation experiment results on matrix multiplication on graphic processing units (GPUs) show that multiple errors are detected at the output in more than 50% of the cases. In the presence of multiple errors, the available hardening strategies may become ineffective or inefficient. Analyzing radiation-induced error distributions, we developed an optimized and experimentally tuned software-based hardening strategy for GPUs. With fault-injection simulations, we compare the performance and correcting capabilities of the proposed technique with the available ones. © 2013 IEEE.
An efficient and experimentally tuned software-based hardening strategy for matrix multiplication on GPUs / Rech, P.; Aguiar, C.; Frost, C.; Carro, L.. - In: IEEE TRANSACTIONS ON NUCLEAR SCIENCE. - ISSN 0018-9499. - 60:4(2013), pp. 2797-2804. [10.1109/TNS.2013.2252625]
An efficient and experimentally tuned software-based hardening strategy for matrix multiplication on GPUs
Rech P.;
2013-01-01
Abstract
Neutron radiation experiment results on matrix multiplication on graphic processing units (GPUs) show that multiple errors are detected at the output in more than 50% of the cases. In the presence of multiple errors, the available hardening strategies may become ineffective or inefficient. Analyzing radiation-induced error distributions, we developed an optimized and experimentally tuned software-based hardening strategy for GPUs. With fault-injection simulations, we compare the performance and correcting capabilities of the proposed technique with the available ones. © 2013 IEEE.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione
