2011 IEEE Ninth International Symposium on Parallel and Distributed Processing With Applications 2011
DOI: 10.1109/ispa.2011.50
|View full text |Cite
|
Sign up to set email alerts
|

Matrix Multiplication on GPUs with On-Line Fault Tolerance

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
22
0

Year Published

2012
2012
2017
2017

Publication Types

Select...
4
2
2

Relationship

1
7

Authors

Journals

citations
Cited by 48 publications
(22 citation statements)
references
References 20 publications
0
22
0
Order By: Relevance
“…Chen [6] analyzes the block row data partitioning scheme for sparse matrices and derives a sufficient condition for recovering critical data without checkpointing. Ding et al [25] construct a column/row checksum matrix for matrix multiplication for GPUs. During computation, the partial product matrix is scanned so that soft errors can be detected and corrected at runtime.…”
Section: B Software Solutions To Soft Errorsmentioning
confidence: 99%
“…Chen [6] analyzes the block row data partitioning scheme for sparse matrices and derives a sufficient condition for recovering critical data without checkpointing. Ding et al [25] construct a column/row checksum matrix for matrix multiplication for GPUs. During computation, the partial product matrix is scanned so that soft errors can be detected and corrected at runtime.…”
Section: B Software Solutions To Soft Errorsmentioning
confidence: 99%
“…The performance results are competitive with those presented by a high-performance GPU, as presented in Ding et al [2011], when considering matrices of the same sizes. The power and energy consumption, on the other hand, are much smaller, even when scaling for the different implementation technologies.…”
Section: Maximum Performance Facing a Memory Wallmentioning
confidence: 81%
“…This section presents a comparison of the RA 3 architecture against the GPU, executing the very efficient ABFT implementation presented in Ding et al [2011]. A GPU is a highly parallel architecture divided into small computing units, named streaming processors, each instantiating a large amount of parallel threads.…”
Section: Comparison Of the Execution Time With The Gpumentioning
confidence: 99%
“…Soft error in the GPU has been exploited [18], and methods have been developed to detect [36,40] and recover from error [35,25,24]. Recently, soft error in matrix multiplication on a GPU has also been studied [10].…”
Section: Related Workmentioning
confidence: 99%