Fault-tolerant iterative solvers with adaptive reliability

Aaditya Shukla; Yue Wu; Saman Zonouz; Maryam Mehri Dehnavi

doi:10.1109/CEFC.2016.7816387

Back

Conference proceeding

Fault-tolerant iterative solvers with adaptive reliability

Aaditya Shukla, Yue Wu, Saman Zonouz and Maryam Mehri Dehnavi

2016 IEEE Conference on Electromagnetic Field Computation (CEFC), pp.1-1

11/2016

DOI: https://doi.org/10.1109/CEFC.2016.7816387

Abstract

Computers

Fault tolerance

Computational modeling

Fault tolerant systems

iterative solvers

reliability

Fault-tolerance

Integrated circuit reliability

Convergence

Soft errors are increasing in modern computers. These faults can corrupt the results of scientific simulations. This work studies error propagation by a bit flip in conjugate gradient (CG) methods. We will also introduce adaptivity to selective reliable fault-tolerant (SRFT) solvers. Our study reduces the compute-intensive reliability steps in SRFT solvers.

Metrics

6 Record Views

Details

Title: Fault-tolerant iterative solvers with adaptive reliability
Creators: Aaditya Shukla - Dept. of Electr. & Comput. Eng., Rutgers Univ., Piscataway, NJ, USA
Yue Wu - Dept. of Electr. & Comput. Eng., Rutgers Univ., Piscataway, NJ, USA
Saman Zonouz - Dept. of Electr. & Comput. Eng., Rutgers Univ., Piscataway, NJ, USA
Maryam Mehri Dehnavi - Dept. of Electr. & Comput. Eng., Rutgers Univ., Piscataway, NJ, USA
Publication Details: 2016 IEEE Conference on Electromagnetic Field Computation (CEFC), pp.1-1
Date published: 11/2016
Publisher: IEEE
Academic Unit: Electrical and Computer Engineering (SOE)
Language: English
Resource Type: Conference proceeding
Identifiers: 991031654484704646