Livermore Computing Division Leader and BlueGene Team Lead discuss progress on the 20-petaFLOPS Sequoia machine.

WCI works with LLNL's Co-Design Team to ensure that the development of the Laboratory's high-performance computing (HPC) systems is informed by the needs of the software developers through a tight collaborative effort early in the design process. Tradeoffs in the hardware and system software designs are informed by the needs of the developers, and the directions of the software research and development community are influenced by early collaborations and insight into those designs.

Co-design is widely considered amongst those with extreme scale computing requirements to be a necessary step for achieving a usable exascale-class system. Application developers must understand, influence, and utilize the changes coming in HPC designs if we are to remain on the rapid performance growth curve Advanced Simulation and Computing (ASC) requires to continue meeting their program mission requirements.

LLNL, ASC, DOE, and the larger HPC community are all facing the following issues:

  • Extreme on-node concurrency
  • Heterogeneous architectures
  • Deep nonuniform memory hierarchies
  • Resilience
  • Power
  • Performance portability
  • Programmer productivity

The Co-Design Team is working on these issues, which are of vital importance for exascale systems, and many of which will also trickle down to the commodity server over the next decade.