CDS&E: Exploiting Multiple Levels of Parallelism in Quantum Chemistry Software

量子化学方法和软件为科学家提供了预测和研究分子系统性质的计算工具。 这些工具被广泛使用，从基础化学和生物化学到药物和材料设计。 这些方法和软件的改进为科学家提供了更强大的发现工具。 该项目将开发量子化学的新方法和软件，旨在充分利用当今计算机硬件的并行能力。 将专门开发单个处理器内核中的并行功能，以及同时使用多个计算节点进行单个计算的能力。 该项目的成功将促进科学的进步，特别是通过减少相同精度的研究所需的计算机时间，并提高使用相同计算时间的研究的准确性，利用量子化学工具有效地研究非常大的分子系统。该项目将首先针对基本上所有量子化学软件中使用的电子排斥积分（ERI）的计算。 使用Obara-Saika（OS）方法计算ERI将被重新组织，以利用现代计算机处理器的单指令多数据（SIMD）功能。 还将讨论OS方法中使用的Boys函数的计算。 将发布用于ERI计算的优化的开源软件库，其中将包括单电子积分和积分导数的功能。 该项目还将进一步开发GTFock量子化学框架，通过添加几个接口使其更易于使用。 GTFock，它具有高效的分布式并行能力，将扩展到包括自适应扰动理论。 GTFock将用于研究大型蛋白质-配体系统。 此外，该项目将涉及本科生，并将用于激励在课堂上的研究。

Quantum chemistry methods and software give scientists a computational tool to predict and study the properties of molecular systems. These tools are used widely, ranging from fundamental chemistry and biochemistry to pharmaceutical and materials design. Improvements in these methods and software give scientists an even more powerful tool for discovery. This project will develop new methods and software for quantum chemistry aimed at fully exploiting the parallel capabilities of today's computer hardware. Parallel capabilities within individual processor cores will be specifically exploited, as well as the ability to use multiple compute nodes concurrently for a single calculation. Success in this project will promote the progress of science, specifically for efficiently studying very large molecular systems with quantum chemistry tools, by decreasing the computer time needed for studies with the same accuracy and increasing the accuracy of studies that use the same computational time. This project will first target the computation of electron repulsion integrals (ERIs) used in essentially all quantum chemistry software. The calculation of ERIs using the Obara-Saika (OS) method will be reorganized to exploit the single instruction multiple data (SIMD) capabilities of modern computer processors. The calculation of the Boys function used in the OS method will also be addressed. An optimized, open source software library for ERI calculation will be released, which will include functionality for one-electron integrals and integral derivatives. This project will also further develop the GTFock quantum chemistry framework to make it easier to use by adding several interfaces. GTFock, which has efficient distributed parallel capabilities, will be extended to include symmetry-adapted perturbation theory. GTFock will be used to study large protein-ligand systems. In addition, this project will involve undergraduate students and will be used to motivate research in the classroom.