The eXascale Programming Environment and System Software (XPRESS) project is part of the DOE Office of Science Advanced Scientific Computing Research X-stack Program initiated in September 2012. The purpose of XPRESS is to devise software that will enable practical extreme-scale computing for DOE missions requiring large-scale computing efforts. Researchers hope to achieve the goal and to have working exascale computing systems available for use by the end of the decade. Shorter-term achievements of XPRESS will include more efficient operation of the DOE’s current petascale computing systems.
XPRESS will define a software architecture to represent the full functionality anticipated for an extreme-scale computing system. It will include interfaces between the programming system and underlying runtime and operating system. Researchers are exploring ways to dramatically improve computing efficiency, including capabilities for higher performing systems and future hardware enhancements. XPRESS directly addresses critical challenges the DOE faces in providing computing infrastructure that is efficient, scalable, and programmable.
RENCI addresses the information flow aspects of the XPRESS software stack. This includes research on how information is passed between layers and exploration of two-way, rather than the current single direction, flow. The work also includes the development of dynamic scheduling environments, and an environment that meets the needs for billion-way parallelism, resilience, reliability, and energy efficiency.
Another RENCI effort under XPRESS looks at optimizing software to take advantage of systems that feature hundreds of thousands of chips that can run millions of concurrent operations. RENCI researchers are studying performance and energy efficiency in these systems with the goal of adaptively managing power and energy, or finding the right balance between the clock speed of chips while minimizing energy consumption. This kind of work is essential to developing extreme-scale systems because although chips are still becoming more powerful, it is too expensive to produce the energy needed to fully power and cool them.
- Allan Porterfield (Principal Investigator)
- Rob Fowler
- Sridutt Bhalachandra
- Indiana University
- Lawrence Berkeley National Laboratory
- Louisiana State University
- Oak Ridge National Laboratory
- Sandia National Laboratory
- University of Houston
- University of Oregon