Tag: exascale computing
In his third column on programming for exascale systems, Michael Wolfe shares his views on what programming at the exascale level is likely to require, and how we can get there from where we are today. He explains that it will take some work, but it’s not a wholesale rewrite of 50 years of high performance expertise.
Is the HPC community too focused on the 10-year milestone?
In Michael Wolfe’s second column on programming for exascale systems, he underscores the importance of exposing parallelism at all levels of design, either explicitly in the program, or implicitly within the compiler. Wolfe calls on developers to express this parallelism, in a language and in the generated code, and to exploit the parallelism, efficiently and effectively, at runtime on the target machine. He reminds the community that the only reason to pursue parallelism is for higher performance.
There are at least two ways exascale computing can go, as exemplified by the top two systems on the latest TOP500 list: Tianhe-1A and Jaguar. The Chinese Tianhe-1A uses 14,000 Intel multicore processors with 7,000 NVIDIA Fermi GPUs as compute accelerators, whereas the American Jaguar Cray XT-5 uses 35,000 AMD 6-core processors.
Exascale computing promises incredible science breakthroughs, but it won’t come easily, and it won’t come free.
Argonne National Laboratory is planning to move up to a 10-petaflop Blue Gene/Q supercomputer next year, supporting the DOE lab’s scientific research. The new machine continues Argonne’s six-year Blue Gene tradition, which has installed every iteration of the architecture in IBM’s BG franchise.
University of Tennessee Professor Jack Dongarra champions exascale computing.
U.S. Department of Energy reveals 2010 INCITE award recipients; and IBM gets closer to exascale with latest advance in on-chip optical communications. We recap those stories and more in our weekly wrapup.
There is a growing feeling that merely taking the latest processor offerings from Intel, AMD or IBM will not get us to exascale in a reasonable time frame, cost budget, and power constraint. One avenue to explore is designing and building more specialized systems, aimed at the types of problems seen in HPC, or at least at the problems seen in some important subset of HPC. Of course, such a strategy loses the advantages we’ve enjoyed over the past two decades of commoditization in HPC; however, a more special purpose design may be wise, or necessary.