Ohio Supercomputer Center parallel programming workshops to give researchers edge with new multi-core processors

COLUMBUS, Ohio (Apr 9, 2007) — 

In preparation for the arrival of a new high performance computing system, the Ohio Supercomputer Center is providing three parallel computing workshops this spring.

The incoming computational system is projected to have 10 times the performance of the largest existing Ohio Supercomputer Center cluster. The IBM e1350 Compute Cluster will have 963 nodes and over 4,000 processor cores, allowing scientists to run their parallelized code on hundreds or thousands of processors.

By attending all three of these workshops, Ohio researchers will have the necessary tools to scale their research on the new computing system. All Ohio academics are invited to attend the workshops, which will be held at the Ohio Supercomputer Center, Columbus, Ohio.

The first workshop, “Parallel Programming with OpenMP,” is being offered April 12 and 13. OpenMP is a de-facto standard for multi-threaded parallel programming on shared-memory multiprocessor systems, based on the use of special comments called compiler directives in the program's source code.

This technique may be of interest to researchers who use current shared memory systems such as the SGI Altix systems, as well those interested in migrating their applications to the new generation of multicore processors. Attendees will learn about the application of OpenMP compiler directives to their codes, optimization strategies to follow, and pitfalls to avoid.

The complementary course to OpenMP, “Parallel Programming with MPI,” will be held May 2 and 3. This workshop focuses on using the Message Passing Interface (MPI) standard to write parallel programs on the Center’s machines.

While OpenMP is useful for parallelizing an application across multiple processor cores within one machine, MPI allows researchers to combine multiple machines connected by a high speed interconnect network to solve larger, more complex problems. MPI topics to be covered include a variety of processor-to-processor communication routines, collective operations performed by groups of processors, defining and using high-level processor connection topologies, and user-specified derived data types for message creation.

Both the OpenMP and MPI workshops will be a combination of lectures and hands-on lab sessions to enhance the learning experience.

Rounding out the training offerings for the quarter, the “Introduction to TotalView” workshop on May 15th will teach attendees how to use the TotalView Debugger. This tool can be used to debug multi-threaded, multi-processor applications. An instructor from TotalView Technologies, Inc. will be teaching the use of the debugger for serial and parallel codes.

“The Ohio Supercomputer Center is excited to offer these workshops to the scientific community in Ohio,” said Jim Giuliani, lead of the Center’s Science and Technology Support group. “A recent poll revealed that a majority of organizations and programmers are not ready to take full advantage of multi-core processors,” Giuliani said. “OSC is taking the steps to prepare Ohio researchers for the arrival of our new system and this new technology to maximize their scientific computing capabilities.”

Registration for the workshops is limited. To register or get more information, write trn-contact@osc.edu, call (614) 292-0890, or visit http://www.osc.edu/supercomputing/notices.

Ohio Supercomputer Center: Celebrating 20 years of service, the Ohio Supercomputer Center is a catalytic partner of Ohio universities and industries that provides a reliable high performance computing and high performance networking infrastructure for a diverse statewide/regional community including education, academic research, industry, and state government. OSC promotes and stimulates computational research and education in order to act as a key enabler for the state's aspirations in advanced technology, information systems, and advanced industries. For additional information, visit http://www.osc.edu.

Subjects: