The main theme of this workshop will be Advanced Parallel Programming and Visualization on Ranger and Spur. TACC instructors will present various advanced topics covering the architecture and proper use of computational and visualization resources, with emphasis on the Sun Constellation Linux Cluster RANGER and the Sun Visualization Cluster SPUR. The class will be of particular interest to computational scientists who are developing code for large clusters and/or porting code to large clusters. The class will provide a foundation for:
- Advanced MPI programming
- Hybrid programming with OpenMP and MPI
- Profiling, optimization and debugging of code
- Advanced visualization techniques
- Visualizing very large datasets
This class will be most suitable for users with experience with UNIX or Linux computer systems, experience with programming, and some basic experience with parallel techniques (familiarity with basic MPI programming).
Participants will have access to RANGER and SPUR at TACC. Linux- Mac- or Windows-based laptops will be needed. To provide full functionality the laptops need to be able to open a secure shell (ssh) with X-forwarding. For the visualization labs a VNC server is needed to allow access to the SPUR desktop and to reduce network traffic. In case that the transatlantic connectivity is bad, provisions for a backup solution have to be made. Two open-source utilities (ParaView and Vislt) will allow to process files, created on SPUR (and copied to the local machines), directly on the laptops. It is advantageous to have the software installed before the training begins.
- VNC server
Dan Stanzione is the Deputy Director of the Texas Advanced Computing Center, and is pleased to be returning to Portugal after teaching at the Summer School in Advanced Computing in 2008 in Coimbra. Dr. Stanzione's research currently focuses on techniques for very large scale debugging of applications on supercomputers, and the construction of cyberinfrastructure to problems in plant biology. Prior to joining TACC, he held positions in advanced computing at Arizona State University and Clemson University, and was Science and Technology Policy Fellow at the National Science Foundation in Washington, DC.
Lars Koesterke joined TACC in the fall of 2007. He is a member of the High Performance Computing (HPC) group, with emphasis on Performance, Evaluation and Optimization of parallel programs. His work focuses on serial optimization, parallel computing (OpenMP and MPI) and Fortran. His background is in Astrophysics (Ph.D. in 1993, Kiel, Germany). Prior to joining TACC he held positions at NASA's Goddard Space Flight Center and at several Universities in the U.S. and in Germany.
Paul Navrátil is a Visualization Scientist in the Visualization and Data Analysis (VDA) group at the Texas Advanced Computing Center. He currently works on algorithms, applications, and system design for large-scale visualization. In addition, Paul is a Ph.D. candidate in Computer Science at the University of Texas at Austin, where he focuses on system-aware optimizations for ray tracing. Prior to joining TACC, Paul held positions at Los Alamos National Laboratory and at Intel Corporation.
|Thursday, May 28|
|09:00||Welcome and Advanced MPI & MPI I/O (slides)||Dan|
|10:30||Lab - Advanced MPI (slides)||All|
|11:30||Hybrid Computing (OpenMP and MPI) (slides)||Lars|
|12:30||Lunch at Círculo Universitário|
|14:00||Incorporating Parallel Libraries in Your Application (slides)||Lars|
|15:00||Advanced Visualization Techniques (slides)||Paul|
|16:30||Lab - Advanced Visualization (slides)||All|
Friday, May 29
|09:00||Profiling and Tuning (mpiP and TAU) (slides)||Lars|
|10:00||Lab - Profiling (slides)||All|
|11:30||Debugging of Parallel Program (slides)||Lars|
|12:30||Lab - Debugging (slides)||All|
|13:00||Lunch at Círculo Universitário|
|14:00||Clouds (Part I) (slides)||Dan|
|15:00||Clouds (Part II) (cont. slides above)||Dan|
|16:30||Dicussion of Possible Future Collaborations||All|