EAGER: CDS&E: Field Programmable Gate Arrays (FPGAs) for Enhancing the Speed and Energy Efficiency of Quantum Chemistry Simulations

点击翻译按钮获取中文摘要

Professor Bryan M. Wong of the University of California-Riverside is supported by an award from the Chemical Theory, Models and Computational Methods program in the Division of Chemistry to utilize new computational hardware, known as Field Programmable Gate Arrays (FPGAs). The goal of this research project is to enhance the speed and energy efficiency of quantum chemistry calculations. Modern quantum chemistry calculations depend critically on massively parallelized computational hardware to enable their predictions. Massively parallel is the term for using a large number of separate computers (or computer processors) to perform a set of coordinated computations simultaneously. Quantitatively accurate predictions from such calculations impact several chemical technologies including combustion, catalysis, process modeling, and chemical production industries. However, the supercomputing centers used to run these predictive calculations consume enormous amounts of energy and resources. Also many such calculations do not scale well as the studied chemical systems grow larger. To address these important issues, FPGAs are being harnessed to enhance the computational speed, as well as to maintain the delicate balance between performance and energy efficiency for large-scale quantum calculations. In addition to addressing these scientific and technological needs, this project is carried out at the University of California-Riverside, which is an accredited Hispanic Serving Institution (HSI). As such, Professor Wong's institution is an ideal place to attract researchers who might otherwise not be aware of the employment opportunities in science and engineering. An FPGA consists of an array of thousands of connection and logic blocks seamlessly connected with each other to create extremely re-configurable programming units. Like Graphics Processing Units (GPUs), FPGAs are ideal for parallelization, but with the added advantage of having reprogrammable circuitry at the hardware level to enable even further parallelization and significant energy gains. This EAGER project is sub-divided into two main (but highly connected) thrusts. Thrust 1 will extend the use of FPGAs to accelerate Fock-matrix builds and Hamiltonian diagonalization for ab initio molecular dynamics (AIMD) and linear-response time-dependent density functional theory calculations, respectively. Thrust 2 will subsequently assess and demonstrate the immense energy efficiency (compared to CPUs/GPUs) of FPGAs for these quantum calculations. The field of quantum chemistry has yet to utilize FPGAs for production-level calculations of any kind, creating an exciting opportunity for transformative leadership in this area. Together these thrusts address basic (yet practical) parallelization issues in quantum chemistry with non-conventional computing approaches, and introduce a new computational capability to perform these quantum chemistry calculations in an energy-efficient manner.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.