Abstract
As the HPC community starts focusing its efforts towards exascale, it becomes clear that we are looking at machines with a billion way concurrency. Although parallel computing has been at the core of the performance gains achieved until now, scaling over 1,000 times the current concurrency can be challenging. As discussed in this paper, even the smallest memory access and synchronization overheads can cause major bottlenecks at this scale. As we develop new software and adapt existing algorithms for exascale, we need to be cognizant of such pitfalls. In this paper, we document our experience with optimizing a fairly common and parallelizable visualization algorithm, threshold of cells based on scalar values, for such highly concurrent architectures. Our experiments help us identify design patterns that can be generalized for other visualization algorithms as well. We discuss our implementation within the Dax toolkit, which is a framework for data analysis and visualization at extreme scale. The Dax toolkit employs the patterns discussed here within the framework's scaffolding to make it easier for algorithm developers to write algorithms without having to worry about such scaling issues.
| Original language | English (US) |
|---|---|
| Title of host publication | Proceedings of SPIE-IS and T Electronic Imaging - Visualization and Data Analysis 2013 |
| Volume | 8654 |
| DOIs | |
| State | Published - Apr 10 2013 |
| Externally published | Yes |
| Event | Visualization and Data Analysis 2013 - Burlingame, CA, United States Duration: Feb 4 2013 → Feb 6 2013 |
Other
| Other | Visualization and Data Analysis 2013 |
|---|---|
| Country | United States |
| City | Burlingame, CA |
| Period | 2/4/13 → 2/6/13 |
Fingerprint
Keywords
- Concurrent Programming
- Programming Techniques
- Software
ASJC Scopus subject areas
- Electronic, Optical and Magnetic Materials
- Condensed Matter Physics
- Computer Science Applications
- Applied Mathematics
- Electrical and Electronic Engineering
Cite this
Optimizing threshold for extreme scale analysis. / Maynard, Robert; Moreland, Kenneth; Atyachit, Utkarsh; Geveci, Berk; Ma, Kwan-Liu.
Proceedings of SPIE-IS and T Electronic Imaging - Visualization and Data Analysis 2013. Vol. 8654 2013. 86540Y.Research output: Chapter in Book/Report/Conference proceeding › Conference contribution
}
TY - GEN
T1 - Optimizing threshold for extreme scale analysis
AU - Maynard, Robert
AU - Moreland, Kenneth
AU - Atyachit, Utkarsh
AU - Geveci, Berk
AU - Ma, Kwan-Liu
PY - 2013/4/10
Y1 - 2013/4/10
N2 - As the HPC community starts focusing its efforts towards exascale, it becomes clear that we are looking at machines with a billion way concurrency. Although parallel computing has been at the core of the performance gains achieved until now, scaling over 1,000 times the current concurrency can be challenging. As discussed in this paper, even the smallest memory access and synchronization overheads can cause major bottlenecks at this scale. As we develop new software and adapt existing algorithms for exascale, we need to be cognizant of such pitfalls. In this paper, we document our experience with optimizing a fairly common and parallelizable visualization algorithm, threshold of cells based on scalar values, for such highly concurrent architectures. Our experiments help us identify design patterns that can be generalized for other visualization algorithms as well. We discuss our implementation within the Dax toolkit, which is a framework for data analysis and visualization at extreme scale. The Dax toolkit employs the patterns discussed here within the framework's scaffolding to make it easier for algorithm developers to write algorithms without having to worry about such scaling issues.
AB - As the HPC community starts focusing its efforts towards exascale, it becomes clear that we are looking at machines with a billion way concurrency. Although parallel computing has been at the core of the performance gains achieved until now, scaling over 1,000 times the current concurrency can be challenging. As discussed in this paper, even the smallest memory access and synchronization overheads can cause major bottlenecks at this scale. As we develop new software and adapt existing algorithms for exascale, we need to be cognizant of such pitfalls. In this paper, we document our experience with optimizing a fairly common and parallelizable visualization algorithm, threshold of cells based on scalar values, for such highly concurrent architectures. Our experiments help us identify design patterns that can be generalized for other visualization algorithms as well. We discuss our implementation within the Dax toolkit, which is a framework for data analysis and visualization at extreme scale. The Dax toolkit employs the patterns discussed here within the framework's scaffolding to make it easier for algorithm developers to write algorithms without having to worry about such scaling issues.
KW - Concurrent Programming
KW - Programming Techniques
KW - Software
UR - http://www.scopus.com/inward/record.url?scp=84875823449&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84875823449&partnerID=8YFLogxK
U2 - 10.1117/12.2007320
DO - 10.1117/12.2007320
M3 - Conference contribution
SN - 9780819494276
VL - 8654
BT - Proceedings of SPIE-IS and T Electronic Imaging - Visualization and Data Analysis 2013
ER -