Abstract
In this paper, we analyze existing MPI benchmarking suites, focusing on two restrictions that prevent them from a wider use in applications and programming systems. The first is a single method of measurement of the execution time of MPI communications implemented by each of the suites. The second one is the design of the suites in the form of a standalone executable program that cannot be easily integrated into applications or programming systems. We present a more flexible benchmarking package, MPIBlib, that provides multiple methods of measurement, both operation-independent and operation-specific. This package can be used not only for benchmarking but also as a library in applications and programming systems for communication performance modeling and optimization of MPI operations. © 2008 Springer-Verlag Berlin Heidelberg.
Author supplied keywords
Cite
CITATION STYLE
Lastovetsky, A., Rychkov, V., & O’Flynn, M. (2008). MPIBlib: Benchmarking MPI communications for parallel computing on homogeneous and heterogeneous clusters. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 5205 LNCS, pp. 227–238). https://doi.org/10.1007/978-3-540-87475-1_32
Register to see more suggestions
Mendeley helps you to discover research relevant for your work.
