Search: WFRF:(Eles Petru Professor)
> (2020-2023) >
Performance and Sec...
Performance and Security Analysis for GPU-Based Applications
-
- Horga, Adrian, 1989- (author)
- Linköpings universitet,Programvara och system,Tekniska fakulteten
-
- Eles, Petru Ion, Professor, 1954- (thesis advisor)
- Linköpings universitet,Programvara och system,Tekniska fakulteten
-
- Peng, Zebo, Professor, 1958- (thesis advisor)
- Linköpings universitet,Programvara och system,Tekniska fakulteten
-
show more...
-
- Rezine, Ahmed, Associate Professor, 1978- (thesis advisor)
- Linköpings universitet,Programvara och system,Tekniska fakulteten
-
- Chattopadhyay, Sudipta, Assistant Professor (thesis advisor)
- Information Systems Technology and Design, Singapore University of Technology and Design
-
- Bombieri, Nicola, Associate Professor (opponent)
- Department of Computer Science, University of Verona, Italy
-
show less...
-
(creator_code:org_t)
- ISBN 9789179294878
- Linköping : Linköping University Electronic Press, 2022
- English 182 s.
-
Series: Linköping Studies in Science and Technology. Dissertations, 0345-7524 ; 2256
- Related links:
-
https://doi.org/10.3...
-
show more...
-
https://liu.diva-por... (primary) (Raw object)
-
https://liu.diva-por... (Preview)
-
https://urn.kb.se/re...
-
https://doi.org/10.3...
-
show less...
Abstract
Subject headings
Close
- Graphics Processing Units (GPUs) are becoming more and more prevalent in general-purpose computing. GPUs are used in areas from embedded systems to super-computing. With applications ranging from fluid dynamics simulations to image processing, machine learning, and encryption, GPU programs need to satisfy not only performance requirements but also various other non-functional constraints. Besides the aspects regarding performance, also security and the worst case execution time (WCET) need to be considered for such GPU applications. In our work, we study such non-functional properties and present approaches to detect and solve issues regarding them.First, we focus on the performance of GPU applications by detecting cache related performance bottlenecks. We detect the root causes of such bottlenecks and provide solutions to reduce their negative impact on performance. We also discuss and compare the impact of cache replacement policies and thread scheduling policies on the performance of GPU applications.Then, we present a measurement-based technique, which combines symbolic execution and genetic algorithms, and is used for estimating the WCET of GPU programs. Our proposed technique helps to produce test inputs that lead towards the WCET of a program. We also propose solutions to alleviate the inherent complexity of GPU programs due to branching behavior and high number of threads running in parallel.In continuation, we propose a technique to expose the side-channel leakage of shared memory in GPU implementations of cryptographic algorithms. We evaluate the robustness of such algorithms in the context of shared memory side-channel leakage. Also, we discuss the security and side-channel leakage for different implementations of the same algorithm.Finally, a formal approach is presented for the detection of GPU shared memory bank conflicts. We explore and discuss the impact of such conflicts on the performance and security of GPU applications. We show how our approach can help in producing inputs that can lead towards the WCET. We also discuss how our approach can be used to evaluate the leakage of the shared memory side-channel for GPU implementations of cryptographic algorithms.
Subject headings
- NATURVETENSKAP -- Data- och informationsvetenskap -- Datavetenskap (hsv//swe)
- NATURAL SCIENCES -- Computer and Information Sciences -- Computer Sciences (hsv//eng)
Publication and Content Type
- vet (subject category)
- dok (subject category)
Find in a library
To the university's database
- By the author/editor
-
Horga, Adrian, 1 ...
-
Eles, Petru Ion, ...
-
Peng, Zebo, Prof ...
-
Rezine, Ahmed, A ...
-
Chattopadhyay, S ...
-
Bombieri, Nicola ...
- About the subject
-
- NATURAL SCIENCES
-
NATURAL SCIENCES
-
and Computer and Inf ...
-
and Computer Science ...
- Parts in the series
-
Linköping Studie ...
- By the university
-
Linköping University