 |
||
|
||
Home > TERATEC FORUM > Workshop 6
Algorithms and parallel software
Transitioning to Many-Core with the MPC Unified Runtime!
Marc PERACHE, CEA,DAM,DIF et Allen MALONY,ParaTools Inc
Download the presentation (Perache)
Download the presentation (Malony)
As the number of cores on a single chip is reaching new records (60 cores, running hundreds of threads), HPC applications are forced to adapt themselves. Not only to take advantage of these new devices but also to remain as efficient than on current architectures. This poses a real problematic for production applications which inherit choices made for previous architectures and requiring to be adapted to new devices. It is well acknowledged that this shift will require model mixing, combining shared and distributed memory parallelism (MPI+OpenMP, MPI+X).
However, efficient model mixing necessarily requires the development of unified runtimes like MPC. In this presentation we present the thread-based MPC runtime, developed by CEA-DAM with support from ParaTools, showing how it has been designed for many-core devices.
We also detail the porting methodology from a pure MPI application to a thread-based one. Process made straightforward thanks to a tight integration in both GNU and Intel compilation chains and allowing performance and memory gains without changing a line of code.
Eventually, we present the tool eco-system which has already been built around MPC, making it eligible for a production use. This includes the TAU Performance System, a performance evaluation tool for MPC and other runtime systems. Then, we conclude with a succinct roadmap of MPC’s development.
 |
Allen.D. Malony is a Professor in the Department of Computer and Information Science at the University of Oregon (UO) where he directs parallel computing research projects, notably the TAU parallel performance system project. He has extensive experience in performance benchmarking and characterization of high-performance computing systems, and has developed performance evaluation tools for a range of parallel machines during the last 25 years. Malony was awarded the NSF National Young Investigator award, was a Fulbright Research Scholar to The Netherlands and Austria, and received the prestigious Alexander von Humboldt Research Award for Senior U.S. Scientists by the Alexander von Humboldt Foundation. He is funded by the Department of Energy, the National Science Foundation, and the Department of Defense. Malony is also the Director of the Neuroinformatics Center at the University of Oregon. He is CEO of ParaTools, Inc. which he founded with Dr. Sameer Shende in 2004. |
 |
Marc Pérache is a research scientist at CEA DAM (Military Applications Department of the French Atomic Energy Authority). He received his PhD in computer science from the University of Bordeaux in 2006 and he joined CEA DAM in 2006. Since 2013, he leads the MPC (Multi-Processor Computing) research team focusing on High-Performance Computing including parallel programming models/runtimes (MPI, Threads, …), high performances networks and memory allocation optimization for multicore and manycore processors. |
Download the presentation (Perache)
Download the presentation (Malony)
- Participation to workshops is free subject to seats available.
- On line registration is obligatory.
- Register now and get your badge here >>>