Hardware Reference
In-Depth Information
MPI rank. For a full description of the problem, please refer to Shende et al.'s
work [5]. Here, TAU can help evaluate an application's I/O performance by
transparently intercepting the I/O calls using a wrapper interposition library.
25.4 Conclusion
Understanding the performance of software packages in combination with
the applications in which they are used requires an ability to capture impor-
tant events and performance data at the library interfaces. This is particu-
larly true for I/O. Here, several techniques for creating wrapper libraries were
presented with the TAU performance system. These techniques were then
demonstrated for tracking I/O that was performed by the GCRM application
on a Cray XE6 system. This work has been instrumental in improving I/O
performance in the GCRM application.
Bibliography
[1] J. Vetter and C. Chambreau. mpiP: Lightweight, Scalable MPI Profiling.
http://mpip.sourceforge.net
, 2014.
[2] K. Lindlan, J. Cuny, A. D. Malony, S. Shende, B. Mohr, R. Rivenburgh,
and C. Rasmussen. A Tool Framework for Static and Dynamic Analysis of
Object-Oriented Software with Templates. In Supercomputing Conference
(SC 2000), 2000.
[3] D. Randall, T. Ringler, R. Heikes, P. Jones, and J. Baumgardner. Cli-
mate Modeling with Spherical Geodesic Grids. Computing in Science and
Engineering, 4:32{41, 2002.
http://kiwi.atmos.colostate.edu/gcrm/
.
[4] S. Shende and A. Malony. The TAU Parallel Performance System. In-
ternational Journal of High Performance Computing Applications, 20(2,
Summer):287{311, 2006. ACTS Collection Special Issue.
[5] S. Shende, A. D. Malony, W. Spear, and K. Schuchardt. Characteriz-
ing I/O Performance Using the TAU Performance System. In D. Bai-
ley, R. Lucas, and S. Williams, editors, Advances in Parallel Computing.
Applications, Tools, and Techniques on the Road to Exascale Computing,
volume 22, pages 647{655. IOS Press, New York, 2010.
[6] T.U. Dresden, ZIH.
VampirTrace.
http://www.tu-dresden.de/zih/
vampirtrace
, 2014.
