Analyzing the instructions vulnerability of dense convolutional network on GPUS

Recently, Deep Neural Networks (DNNs) have been increasingly deployed in various healthcare applications, which are considered safety-critical applications. Thus, the reliability of these DNN models should be remarkably high, because even a small error in healthcare applications can lead to injury o...

Full description

Bibliographic Details
Main Authors: Khalid, Adam, Izzeldin, I. Mohd, Ibrahim, Younis
Format: Article
Language:English
Published: Institute of Advanced Engineering and Science 2021
Subjects:
Online Access:http://umpir.ump.edu.my/id/eprint/30696/
http://umpir.ump.edu.my/id/eprint/30696/1/Analyzing%20the%20instructions%20vulnerability%20of%20dense%20convolutional%20network%20on%20GPUS.pdf
_version_ 1848823578066157568
author Khalid, Adam
Izzeldin, I. Mohd
Ibrahim, Younis
author_facet Khalid, Adam
Izzeldin, I. Mohd
Ibrahim, Younis
author_sort Khalid, Adam
building UMP Institutional Repository
collection Online Access
description Recently, Deep Neural Networks (DNNs) have been increasingly deployed in various healthcare applications, which are considered safety-critical applications. Thus, the reliability of these DNN models should be remarkably high, because even a small error in healthcare applications can lead to injury or death. Due to the high computations of the DNN models, DNNs are often executed on the Graphics Processing Units (GPUs). However, the GPUs have been reportedly impacted by soft errors, which are extremely serious issues in the healthcare applications. In this paper, we show how the fault injection can provide a deeper understanding of DenseNet201 model instructions vulnerability on the GPU. Then, we analyze vulnerable instructions of the DenseNet201 on the GPU. Our results show that the most significant vulnerable instructions against soft errors PR, STORE, FADD, FFMA, SETP and LD can be reduced from 4.42% to 0.14% of injected faults, after we applied our mitigation strategy.
first_indexed 2025-11-15T02:59:21Z
format Article
id ump-30696
institution Universiti Malaysia Pahang
institution_category Local University
language English
last_indexed 2025-11-15T02:59:21Z
publishDate 2021
publisher Institute of Advanced Engineering and Science
recordtype eprints
repository_type Digital Repository
spelling ump-306962021-10-14T02:18:38Z http://umpir.ump.edu.my/id/eprint/30696/ Analyzing the instructions vulnerability of dense convolutional network on GPUS Khalid, Adam Izzeldin, I. Mohd Ibrahim, Younis QA76 Computer software TA Engineering (General). Civil engineering (General) Recently, Deep Neural Networks (DNNs) have been increasingly deployed in various healthcare applications, which are considered safety-critical applications. Thus, the reliability of these DNN models should be remarkably high, because even a small error in healthcare applications can lead to injury or death. Due to the high computations of the DNN models, DNNs are often executed on the Graphics Processing Units (GPUs). However, the GPUs have been reportedly impacted by soft errors, which are extremely serious issues in the healthcare applications. In this paper, we show how the fault injection can provide a deeper understanding of DenseNet201 model instructions vulnerability on the GPU. Then, we analyze vulnerable instructions of the DenseNet201 on the GPU. Our results show that the most significant vulnerable instructions against soft errors PR, STORE, FADD, FFMA, SETP and LD can be reduced from 4.42% to 0.14% of injected faults, after we applied our mitigation strategy. Institute of Advanced Engineering and Science 2021-10 Article PeerReviewed pdf en cc_by_sa_4 http://umpir.ump.edu.my/id/eprint/30696/1/Analyzing%20the%20instructions%20vulnerability%20of%20dense%20convolutional%20network%20on%20GPUS.pdf Khalid, Adam and Izzeldin, I. Mohd and Ibrahim, Younis (2021) Analyzing the instructions vulnerability of dense convolutional network on GPUS. International Journal of Electrical and Computer Engineering (IJECE), 11 (5). pp. 4481-4488. ISSN 2088-8708 (Print); 2722-2578 (Online). (Published) http://ijece.iaescore.com/index.php/IJECE/article/view/24607/15136 http://doi.org/10.11591/ijece.v11i5.pp4481-4488
spellingShingle QA76 Computer software
TA Engineering (General). Civil engineering (General)
Khalid, Adam
Izzeldin, I. Mohd
Ibrahim, Younis
Analyzing the instructions vulnerability of dense convolutional network on GPUS
title Analyzing the instructions vulnerability of dense convolutional network on GPUS
title_full Analyzing the instructions vulnerability of dense convolutional network on GPUS
title_fullStr Analyzing the instructions vulnerability of dense convolutional network on GPUS
title_full_unstemmed Analyzing the instructions vulnerability of dense convolutional network on GPUS
title_short Analyzing the instructions vulnerability of dense convolutional network on GPUS
title_sort analyzing the instructions vulnerability of dense convolutional network on gpus
topic QA76 Computer software
TA Engineering (General). Civil engineering (General)
url http://umpir.ump.edu.my/id/eprint/30696/
http://umpir.ump.edu.my/id/eprint/30696/
http://umpir.ump.edu.my/id/eprint/30696/
http://umpir.ump.edu.my/id/eprint/30696/1/Analyzing%20the%20instructions%20vulnerability%20of%20dense%20convolutional%20network%20on%20GPUS.pdf