Evaluation and Mitigation of Radiation-Induced Soft Errors in Graphics Processing Units

Oliveira, Daniel; Pilla, Laércio Lima; Santini, Thiago; Rech, Paolo

doi:10.1109/tc.2015.2444855

Cited by 85 publications

(75 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The main causes of transient errors are voltage-frequency variations, temperature perturbations, and electromagnetic interferences. Lately, neutron-induced errors have been shown to be critical for HPC systems [15,46]. A flux of about 13 neutrons/((cm 2 ) × h) reaches ground at sea level, and the flux exponentially increases with altitude [29].…”

Section: Background 21 Transient Errors Effects In Hpcmentioning

confidence: 99%

“…Highly parallel computing architectures, like the Xeon Phi, have some reliability weaknesses [15,16,21,49]. For instance, a single particle generating a radiation-induced failure in the scheduler or shared memories (used to expedite parallel executions), is likely to affect the computation of several parallel threads.…”

Section: Background 21 Transient Errors Effects In Hpcmentioning

confidence: 99%

See 1 more Smart Citation

Experimental and analytical study of Xeon Phi reliability

Oliveira

Pilla

DeBardeleben

et al. 2017

Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis

Self Cite

View full text Add to dashboard Cite

We present an in-depth analysis of transient faults effects on HPC applications in Intel Xeon Phi processors based on radiation experiments and high-level fault injection. Besides measuring the realistic error rates of Xeon Phi, we quantify Silent Data Corruption (SDCs) by correlating the distribution of corrupted elements in the output to the application's characteristics. We evaluate the benefits of imprecise computing for reducing the programs' error rate. For example, for HotSpot a 0.5% tolerance in the output value reduces the error rate by 85%.We inject different fault models to analyze the sensitivity of given applications. We show that portions of applications can be graded by different criticalities. For example, faults occurring in the middle of LUD execution, or in the Sort and Tree portions of CLAMR, are more critical than the remaining portions. Mitigation techniques can then be relaxed or hardened based on the criticality of the particular portions.

show abstract

Section: Background 21 Transient Errors Effects In Hpcmentioning

confidence: 99%

Section: Background 21 Transient Errors Effects In Hpcmentioning

confidence: 99%

Experimental and analytical study of Xeon Phi reliability

Oliveira

Pilla

DeBardeleben

et al. 2017

Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis

Self Cite

View full text Add to dashboard Cite

show abstract

“…It is worth noting that a significant component of Crashes is caused by radiation corruption of the device control circuitry. Errors affecting instruction memory, the GPU hardware schedulers, or the CPU-GPU interface could lead to application crash or system hang, independently of the algorithm proprieties [de Oliveira et al 2016]. Crashes, while more frequent than SDCs in HOG, are considered less critical as they are easily detected [Li et al 2008;Nakka et al 2005;Pattabiraman et al 2006].…”

Section: Crashmentioning

confidence: 99%

“…While extremely efficient in terms of FLOP/s (floating-point operations per second) and FLOPs-per-WATT, modern GPUs have been shown to be prone to experience radiation-induced corruption [DeBardeleben et al 2013;Wunderlich et al 2013;Gomez et al 2014;Oliveira et al 2014;de Oliveira et al 2016]. GPU architecture may be particularly susceptible to be corrupted by radiation for three main reasons.…”

Section: Introductionmentioning

confidence: 99%

“…As shown in this article, the lack of architectural hardening makes embedded GPUs prone to experience soft errors when applied in safety-critical systems. It is worth noting that the latest GPUs for High Performance Computing applications include ECC on the main memory structures [NVIDIA 2015;Intel 2014], which has been shown to reduce of about one order of magnitude the GPU error rate [de Oliveira et al 2016]. Lately, some SoCs with ECC-protected GPU memories have appeared in the market [Nvidia 2016;AMD 2015] but are not employed in any self-driving car yet.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Evaluation of Histogram of Oriented Gradients Soft Errors Criticality for Automotive Applications

Fernandes

Weigel

Jung

et al. 2016

ACM Trans. Archit. Code Optim.

Self Cite

View full text Add to dashboard Cite

Pedestrian detection reliability is a key problem for autonomous or aided driving, and methods that use Histogram of Oriented Gradients (HOG) are very popular. Embedded Graphics Processing Units (GPUs) are exploited to run HOG in a very efficient manner. Unfortunately, GPUs architecture has been shown to be particularly vulnerable to radiation-induced failures. This article presents an experimental evaluation and analytical study of HOG reliability. We aim at quantifying and qualifying the radiation-induced errors on pedestrian detection applications executed in embedded GPUs. We analyze experimental results obtained executing HOG on embedded GPUs from two different vendors, exposed for about 100 hours to a controlled neutron beam at Los Alamos National Laboratory. We consider the number and position of detected objects as well as precision and recall to discriminate critical erroneous computations. The reported analysis shows that, while being intrinsically resilient (65% to 85% of output errors only slightly impact detection), HOG experienced some particularly critical errors that could result in undetected pedestrians or unnecessary vehicle stops. Additionally, we perform a fault-injection campaign to identify HOG critical procedures. We observe that Resize and Normalize are the most sensitive and critical phases, as about 20% of injections generate an output error that significantly impacts HOG detection. With our insights, we are able to find those limited portions of HOG that, if hardened, are more likely to increase reliability without introducing unnecessary overhead.

show abstract

Microprocessor Testing

Quinn¹

2019

Radiation Effects on Integrated Circuits and Systems for Space Applications

View full text Add to dashboard Cite

Evaluation and Mitigation of Radiation-Induced Soft Errors in Graphics Processing Units

Cited by 85 publications

References 29 publications

Experimental and analytical study of Xeon Phi reliability

Experimental and analytical study of Xeon Phi reliability

Evaluation of Histogram of Oriented Gradients Soft Errors Criticality for Automotive Applications

Microprocessor Testing

Contact Info

Product

Resources

About