Computational science is firmly established as a pillar of scientific discovery and technology, promising unprecedented new capabilities. The National Strategic Computing Initiative (NSCI) establishes an ambitious roadmap to advance Science and Technology (S&T) through support for sustained innovations in high performance computing and its use. Harnessing the power of millions of computer cores and/or compute accelerators (also called graphical processing units, GPUs) foreseen in future high performance computers requires a new generation of application software and algorithms, able to effectively utilize such resources and create the revolutionary S&T advances that underpin the nation's economic competitiveness. The proposed work will develop high-performance, scalable quantum simulation software for complex materials and devices, which will be portable and tunable across alternative exascale architectures. It will be able to address grand challenges in the design of quantum materials and devices, responding to one of the five strategic objectives of NSCI. Quantum materials and processes also underpin one of NSF's 10 Big Ideas for Future NSF Investments, The Quantum Leap: Leading the Next Quantum Revolution. Another important program in which materials simulation plays a key role is the Materials Genome Initiative. It seeks to "deploy advanced materials at least twice as fast at a fraction of the cost" and relies on computational materials design as the critical aspect, with computation guiding experiments. The goals of this project are to refactor and extend the open-source RMG software suite to future computer architectures at exascale, to enable transformational research on the design of quantum materials and devices from fundamental quantum-mechanical level. The RMG software will extend from desktops to the largest supercomputer systems, and will also perform well on a multitude of other systems, such as parallel computing clusters of various sizes, including those with GPUs. At the highest level of performance, it will enable predictive simulations at unprecedented scale, impact several areas of science and engineering and become a source of new discoveries and economic growth. RMG, already highly parallel and capable of multi-petaflops speeds, can provide a pathway towards reaching key NSCI goals. RMG has already been included in a benchmark suite which will be used to help select future supercomputers. At the same time, it's scalability means that it will be useful in classroom education running on students' laptops, to help individual researchers perform significant scientific or technological research on their accelerator- or GPU-equipped workstations, and, to run larger problems on a multitude of computer clusters with varying capabilities.
The goals of this project are to refactor and extend the open-source RMG software suite to exascale architectures, to enable transformational research on the design of quantum materials and devices from fundamental quantum-mechanical level. The RMG software will extend from desktops to the largest supercomputer systems, and will also perform well on a multitude of other systems, such as parallel clusters of various sizes, including those with GPUs. At the highest level of performance, it will enable predictive simulations at unprecedented scale, impact several areas of science and engineering and become a source of new discoveries and economic growth. RMG, already highly parallel and capable of multi-petaflops speeds, can provide a pathway towards reaching some of key NSCI goals. It has been included just as a part of NSF's Sustained Petascale Performance Benchmarks, which will be used to select NSF's future Leadership Class supercomputers. However, it will also be useful in classroom education, running on individual students' laptops, help individual researchers perform significant scientific or technological research on their accelerator- or GPU-equipped workstations, and also run on a multitude of clusters with varying capabilities. The extensible and portable exascale-capable software tools for simulations of complex quantum materials and devices will enable many scientific and technological endeavors that are currently too difficult to pursue, including dramatically accelerated discovery and design of complex quantum materials structures, such as nanostructured energy storage materials; nanoscale biosensors for electrical sequencing of DNA and nanoscale "laboratories on a chip" for monitoring health; as well as addressing fundamental questions about quantum behavior and the manipulation of quantum systems. Analogous accelerated progress is expected in other areas of science and technology that depend on nano and meso scales that are intermediate between those of molecules and bulk solids. Medium-size simulations will be enabled on local computing platforms, with an easy migration pathway to national facilities with the same input GUI. The exascale quantum simulation software will thus become a major resource to the national community. The easy availability of desktop binaries, supported source code, and optimized binaries at national facilities will lead to a major increase in high-end usage, dramatically enlarging the number and quality of simulations. The increase in users at all levels will stimulate their contributions both by new development and though incorporation of existing code elements into various materials frameworks. The national Cyberinfrastructure Community will be engaged through SI2 Software Institutes, Blue Waters and XSEDE projects, including live tutorials at workshops, as well tutorial sessions at conferences. STEM education and interests will be addressed by recruitment of undergraduate students, visually attractive presentations at libraries and science museums, and web-based presentation modules.
This project is supported by the Office of Advanced Cyberinfrastructure in the Directorate for Computer & Information Science and Engineering and the Division of Materials Research in the Directorate of Mathematical and Physical Sciences.
