The project updates the GTC-Princeton (GTC-P) code, which was originally developed for fusion research applications at PPPL and has evolved into highly portable software that is deployed on supercomputers worldwide. The National Science Foundation (NSF) strongly supported advances in the code from 2011 through 2014 through the "G8" international extreme scale computing programme, which represented the United States and seven other highly industrialized countries during that period.
Heading the new IPCC activity for the University's Princeton Institute for Computational Science & Engineering (PICSciE) is William Tang, a PPPL physicist and PICSciE principal investigator (PI). Working with William Tang is Co-PI Bei Wang, Associate Research Scholar at PICSciE, who leads this accelerated modernization effort. Joining them in the project are Co-PIs Carlos Rosales of the NSF's Texas Advanced Computing Center at the University of Texas at Austin and Khaled Ibrahim of the Lawrence Berkeley National Laboratory.
The current GTC-P code has advanced understanding of turbulence and confinement of the superhot plasma that fuels fusion reactions in doughnut-shaped facilities called tokamaks. Understanding and controlling fusion fuel turbulence is a grand challenge of fusion science, and great progress has been made in recent years. It can determine how effectively a fusion reactor will contain energy generated by fusion reactions, and thus can strongly influence the eventual economic attractiveness of a fusion energy system. Further progress on the code will enable researchers to study conditions that arise as tokamaks increase in size to the enlarged dimensions of ITER - the flagship international fusion experiment under construction in France.
Through the IPCC, Intel will provide access to systems for exploring the modernization of the code. Included will be clusters equipped with the most recent Intel "Knights Landing" (KNL) central processing chips.
The upgrade will become part of the parent GTC code, which is led by Professor Zhihong Lin of the University of California, Irvine, with William Tang as co-PI. That code is also being modernized and will be proposed, together with GTC-P, to be included in the early science portfolio for the Aurora supercomputer. Aurora will begin operations at the Argonne Leadership Computing Facility, a DOE Office of Science User Facility at Argonne National Laboratory, in 2019. Powering Aurora will be Intel "Knights Hill" processing chips.
Last year, the GTC and GTC-P codes were selected to be developed as an early science project designed for the Summit supercomputer that will be deployed at Oak Ridge Leadership Computing Facility, also a DOE Office of Science User Facility, at Oak Ridge National Laboratory in 2018. That modernization project differs from the one to be proposed for Aurora because Summit is being built around architecture powered by NVIDIA Volta graphical processing units and IBM Power 9 central processing chips.
Moreover, the code planned for Summit will be designed to run on the Aurora platform as well.
The two new machines will boost U.S. computing power far beyond Titan, the current leading U.S. supercomputer at Oak Ridge that can perform 27 quadrillion - or million billion - calculations per second. Summit and Aurora plan to perform some 200 quadrillion and 180 quadrillion calculations per second, respectively. Stated William Tang: "These new machines hold tremendous promise for helping to accelerate scientific discovery in many application domains, including fusion, that are of vital importance to the country."