Short Course Announcement:
Parallel Programming on Linux Beowulf Clusters
Instructor: Dr. Dan Stanzione
The Fulton High Performance Computing Initiative, back by popular demand, would like to announce this short course intended to provide hands-on training in the use of Beowulf cluster parallel computers. This course is open to all students, staff, postdocs and faculty at Arizona State University on a first-come, first-serve basis, and will be repeated periodically as needed.
Each of the courses includes a ~90 minute lecture, followed by a hands-on lab session. The lecture/lab sessions will run 1:30-5:00 each day.
Examples will be presented in both FORTRAN and C. (Space available on a first come, first serve basis) Covers the basics of Beowulf philosophy, architecture, and parallel computing models. Tools to access the system, run and monitor jobs. Introductory MPI and OpenMP programming, simple examples. July 2nd, 2007, 1:30 – 5:00, ERC Rm 490 July 23rd, 2007, 1:30 – 5:00, ERC Rm 490 Covers the incorporation of parallel I/O into parallel programming for large dataset computation. Out-of-core solution techniques. The MPI-I/O interface and ROMIO implementation. In addition, in this final session we will explore alternatives to MPI, including OpenMP and the CxC parallel language.
June 25th, 2007, 1:30 – 5:00, ERC Rm 490
#1 Beowulfs, Mini Grids, and Basic MPI
#2 Advanced MPI Programming
Covers the rest of the MPI interface, collective operations (reduce, scatter, gather), use of communicators, user defined datatypes for advanced data structure decomposition. Example applications in linear algebra problems.
July 16th, 2007, 1:30 – 5:00, ERC Rm 490
#3 Parallel Algorithms
A brief look at common problems and their solutions in parallel computing. Pipeline, processor farm, and data parallel models of parallel program. Analysis of data dependencies in loops. Examine issues of efficiency, speedup, isoefficiency, and granularity. Example problems, good and bad parallel decompositions.
#4 Parallel I/O , Debugging, and OpenMP