FPGA-based High Performance Computing

描述(由申请人提供)：对更具成本效益、更灵活、更方便的高性能计算的需求是生物信息学和计算生物学不同领域的共同主线。我们通过使用基于可配置电路或现场可编程门阵列的新一代计算机来解决这个问题，这种计算机能够使每个节点的速度比当前基于微处理器的系统提高100倍。要开发的主要技术涉及从三个关键领域创建FGPA加速版本的生产应用程序：分子动力学、序列比对和对接。主要目标是使这些应用程序在经济高效、易于获得的平台上运行；扩展到大型系统；经过验证并对最终用户透明；对于硬件、软件和算法的更改具有经济上的可维护性；并分阶段交付，在授权期早期生产有用的人工制品，然后在之后定期交付。对这些目标，尤其是可维护性来说，重要的是继续开发支持设计的基础设施。潜在的意义在于当前任务的更高吞吐量、更高的结果质量、更广泛地使用高性能计算，或者完全能够解决更雄心勃勃的问题。拟议工作的创新分为两类：应用程序开发中使用的方法，包括为执行这些方法而创建的设计工具；以及每个应用程序、现场可编程门阵列感知、重组和算法。前者解决的一个关键问题是应用专家和FPGA/逻辑设计者之间的分工，这是创建高效、生产质量的生命科学应用所必需的。与公共健康相关的是提高生命科学研究人员的生产率：增强的计算能力导致更多的假设得到更详细的探索，或者在给定的周转时间内处理更复杂的现象。例如，这将允许对细胞的多个部分一起建模，而不是一次一个，从而增加对疾病过程的理解，并对改进药物设计产生影响。

DESCRIPTION (provided by applicant): The need for more cost-effective, flexible, and convenient high-performance computing is a common thread across diverse areas of Bioinformatics and Computational Biology. We address this problem by enabling the use of a new generation of computers, based on configurable circuits or FPGAs, that is capable of delivering 100-fold speed-ups per node over current microprocessor-based systems. The primary technology to be developed involves creating FGPA-accelerated versions of production applications from three key domains: molecular dynamics, sequence alignment, and docking. The primary aims are for these applications to run on cost-effective, readily available, platforms; scale to large systems; be validated and transparent to the end user; be economically maintainable with respect to changes in hardware, software, and algorithm; and be delivered in stages, with useful artifacts being produced early in the grant period and then at regular intervals thereafter. Important to these goals, especially maintainability, is the continued development of supporting design infrastructure. The potential significance lies in higher throughput of current tasks, higher quality of results, wider use of high performance computing, or the capability of addressing more ambitious problems altogether. The innovation of the proposed work falls into two categories: the methods used in application development, including the design tools created to execute those methods; and the per application, FPGA-aware, restructurings and algorithms. A key issue addressed in the former is the division of labor among application specialist and FPGA/logic designer that is required to create highly efficient, production-quality, life science applications. The relevance to public health is in improving the productivity of the life science researcher: enhanced computing capability leads to more hypotheses explored in finer detail, or the processing of more complex phenomena within a given turnaround time. For example, this would allow multiple parts of a cell to be modeled together, rather than one at a time, resulting in increased understanding of disease processes with implications for improved drug design.