Self-Tuned Adaptive Routines for MPI Collective Operations (STAR-MPI)
[SourceForge Summary Page]
The STAR-MPI (Self-Tuned Adaptive Routines for MPI Collective Operations) project
is an ongoing research and development effort at Florida State University focusing
on applying empirical adaptive techniques to supporting portable performance for MPI
collective operations. At present, it
provides C and Fortran77 interfaces to five MPI collective routines:
MPI_Alltoall, MPI_Allreduce, MPI_Allgather, MPI_Bcast, and MPI_Reduce.
The following two papers describe the technical details and background
of this package.
- A. Faraj, X. Yuan, and D.K. Lowenthal,
Tuned Adaptive Routines for MPI Collective Operations," the 20th ACM
International Conference on Supercomputing, pages 199-208, Cairns, Australia,
- A. Faraj, P. Patarasuk, and X. Yuan,
"A Study of Process Arrival Patterns for MPI Collective Operations," International
Journal of Parallel Programming, 2008.
You can download STAR-MPI
Please see the README file in the package for installation and usage.
Please use SourceForge support and bug tracker at
the SourceForge summary page if you
need any help.
This project is supported in part by National Science Foundation under grants CCF-0342540,
CCF-0541096, and CNS-0551555.