Talk Title:
Performance Challenges for MPI Implementations

Abstract:
The Message Passing Interface (MPI) is the dominant parallel programming model for technical computing. One of the reasons for its success is that it makes it possible for programmers to achieve high performance in their applications. However, not all features of MPI are efficiently used or implemented, particularly those features that were introduced in the MPI-2 standard. Micro-benchmarks, such as the various "ping pong" tests, are often used to measure and tune MPI implementations, but these neither reflect common applications use nor exploit many of the features of MPI. Application benchmarks rarely cover the different ways in which MPI may be used.

In this talk, I describe some small benchmarks that were motivated by typical application uses of MPI for point-to-point communication, one-sided communication, multithreaded MPI, and collective communication.

Measurements of the performance several MPI implementations, including MPICH2 and OpenMPI as well as those from IBM, SGI, and Sun, on these benchmarks will be discussed. These results point out several areas where new approaches are needed, particularly for support of multi-threaded MPI applications on the next generation of multi- and many-core processors.

Bio:
William Gropp received his B.S. in Mathematics from Case Western Reserve University in 1977, a MS in Physics from the University of Washington in 1978, and a Ph.D. in Computer Science from Stanford in 1982. He held the positions of assistant (1982-1988) and associate (1988-1990) professor in the Computer Science Department at Yale University. In 1990, he joined the Numerical Analysis group at Argonne, where he was a Senior Computer Scientist in the Mathematics and Computer Science Division, a Senior Scientist in the Department of Computer Science at the University of Chicago, and a Senior Fellow in the Argonne-Chicago Computation Institute. His research interests are in parallel computing, software for scientific computing, and numerical methods for partial differential equations. He has played a major role in the development of the MPI message-passing standard. He is co-author of the most widely used implementation of MPI, MPICH, and was involved in the MPI Forum as a chapter author for both MPI-1 and MPI-2. He has written many books and papers on MPI including "Using MPI" and "Using MPI-2". He is also one of the designers of the PETSc parallel numerical library, and has developed efficient and scalable parallel algorithms for the solution of linear and nonlinear equations. Gropp was named an ACM Fellow in 2006. He recently joined the University of Illinois at Urbana-Champaign as a professor in Computer Science.