Hardware Reference
In-Depth Information
and exchange diverse scientific datasets. As a serial library, Silo has nonethe-
less demonstrated scalable performance for many HPC applications (ALE3D,
Kull, Overlink, PMesh, Ares, and VisIt are some examples) using the MIF-IO
parallel I/O approach. MIF-IO has demonstrated good performance to 100,000
tasks and is currently in process of being scaled to 1,000,000 cores. MIF-IO
has a number of advantages over SSF-IO making its application very attrac-
tive in HPC workflows involving diverse and disparate datasets with dramatic
variation in size, shape, and existence of data across MPI tasks. Silo's success
is due in large part to the performance and capabilities available in HDF5.
Bibliography
[1] ASC Dawn Machine.
https://computing.llnl.gov/tutorials/bgp/
#Hardware
, 2010.
[2] LLNL Graph Machine.
https://computing.llnl.gov/?set=
resources&page=SCF_resources#graph
, 2012.
[3] ASCI Purple Machine.
https://asc.llnl.gov/computing_resources/
purple/
, 2010.
[4] Silo User's Guide: Version 4.7.1.
https://e-reports-int.llnl.gov/
pdf/382967.pdf
, 2009.
[5] Marc Snir, Steve Otto, Steven Huss-Lederman, David Walker, and Jack
Dongarra. MPI-The Complete Reference, Volume 1: The MPI Core. MIT
Press, Cambridge, MA, USA, 2nd. (revised) edition, 1998.
[6] Bethel W., Childs H., and Hansen C. High Performance Visualization:
Enabling Extreme-Scale Scientific Insight. Chapman and Hall/CRC, 2012.
[7] Futral W.S., Dube E., Neely J.R., and Pierce T.G. Performance of ALE3D
on the ASCI Machines. In Nuclear Explosives Code Development Confer-
ence, 1999.
https://e-reports-ext.llnl.gov/pdf/234827.pdf
.
