NERSC Selects 20 NESAP Code Teams

The National Energy Research Scientific Computing Center (NERSC) has accepted 20 projects into the NERSC Exascale Scientific Applications Program (NESAP), a new collaborative effort that partners NERSC, Intel, and Cray resources with code teams across the US to prepare for Cori, the center’s next-generation supercomputer.

Cori, a Cray XC system slated to be deployed at NERSC in 2016, is intended to meet the growing computational needs of the Department of Energy’s (DOE) science community and serve as a platform for transitioning users to energy-efficient, manycore architectures.

NESAP represents a unique opportunity for researchers to work with NERSC and vendor staff to optimize application codes for Cori's Knights Landing manycore architecture, which may represent an approach to exascale systems, according to Harvey Wasserman, HPC consultant at NERSC and NESAP post-doc lead. The program will provide access to early hardware and special training and preparation sessions, plus an opportunity for a postdoctoral researcher to investigate computational science issues associated with energy-efficient manycore systems.

“We were very impressed with the response to our call for participation – we received 50 proposals,” Wasserman said. “NERSC is very grateful for the user participation; they’ll be doing the ‘heavy lifting’ during the project and will help us ensure that the workload is ready when Cori is deployed. This exciting machine architecture is now being followed by exciting science in the national interest.”

Broad Range of Science

The selected projects were chosen based on computational and scientific reviews by NERSC and other DOE staff. The projects span a range of scientific fields – including astrophysics, genomics, materials science, climate and weather modeling, plasma fusion physics and accelerator science – and represent a large portion of NERSC’s current and projected computational workload. A variety of numerical methods and computational approaches are also represented, including particle-mesh algorithms, adaptive mesh refinement techniques, molecular dynamics, eigenvalue problems for complex molecular systems, and ab initio computational chemistry methods. In addition, several codes are directly involved in DOE Scientific Discovery through Advanced Computing (SciDAC) activities.

“We didn’t just choose codes that we knew would be relatively easy to run on Cori," said Katie Antypas, NERSC’s Services Department Head and Cori principal investigator. “NERSC has an obligation to all DOE Office of Science projects, and that meant choosing some challenging codes. With assistance from Cray, Intel, and the HPC scientific library software community, we’re certain these codes will be successful on Cori.”

Antypas also emphasized that NESAP is about early access to hardware and that all 44 projects will be able to get on Cori as soon as it’s available at NERSC.

As part of NESAP, NERSC will soon advertise for eight postdoctoral scholars who will be hired to research computational and computer science issues associated with optimization techniques for application codes on manycore architectures. Which of the 20 selected projects will be assigned a postdoc has not yet been decided and may depend on the expertise and career interests of the applicants.