IRES: Track I: Collaborative Research: Supporting FSU and MTU Student Research with NTNU Faculty on Automatic Improvement of Application Performance

随着计算机技术的进步和程序执行速度的加快，更多的计算机应用成为可能。该项目将使佛罗里达州立大学（FSU）和密歇根理工大学（MTU）的学生能够访问挪威科技大学（NTNU）进行研究，从而使计算机应用程序的性能和能源效率得到自动提高。在这个项目的三年里，16名FSU和MTU学生（每组5人、5人、6人）将在5月、6月和7月访问位于挪威特隆赫姆的NTNU，为期10周。在此期间，学生将与NTNU计算机体系结构实验室（CAL）附属的Magnus Sjalander和Magnus Jahre教授的研究小组中的教职员工，博士后助理和研究生密切合作。参加FSU和MTU的学生不仅会增加他们的研究知识，而且还会变得更加全球化，并为在多元文化的国际环境中工作做好更好的准备。在计算机架构或编译器没有创新的情况下提高处理器性能的时代已经结束，因为近年来由于热限制而无法提高计算机的时钟速率。然而，手动修改程序以有效地利用计算机体系结构是耗时的，容易出错的，并且不能移植到其他系统。提高应用程序性能的最有效方法是在没有应用程序开发人员干预的情况下自动利用体系结构特性。我们的重点将是通过生成代码来利用机器指令级别的现有和拟议的架构特性，自动实现高性能和节能。我们建议开发编译工具，以促进自动生成代码的过程，以利用这些拟议的架构特性，并开发仿真工具来评估对性能和能源效率的影响。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

When the technology for computers advance and programs execute faster, more computer applications become possible. This project will enable Florida State University (FSU) and Michigan Technology University (MTU) students to visit the Norwegian University of Science and Technology (NTNU) to conduct research that will allow the performance and energy efficiency of computer applications to be automatically improved. Over the three years of this project, 16 FSU and MTU students (in cohorts of five, five, and six) will visit NTNU, which is in Trondheim, Norway, for a period of 10 weeks during May, June, and July. While in residence, the students will work closely with the faculty, postdoctoral associates, and graduate students in the research groups of Professors Magnus Sjalander and Magnus Jahre who are affiliated with the Computer Architecture Laboratory (CAL) at NTNU. The participating FSU and MTU students will not only increase their research knowledge, but will also become more globally engaged and better prepared to work in a culturally diverse, international environment. The era of improving processor performance without innovations in computer architecture or compilers is over since increasing the clock rate for computers has not been possible in recent years due to thermal limitations. However, manually modifying programs to efficiently exploit computer architectures is time consuming, error prone, and not portable to other systems. The most effective way to improve application performance is to automatically exploit architectural features without the intervention of the application developers. Our focus will be on automatically achieving high performance and energy efficiency by generating code to exploit existing and proposed architectural features at the machine instruction level. We propose to develop the compilation tools to facilitate the process of automatically generating code to exploit these proposed architectural features and to develop the simulation tools to evaluate the impact on both performance and energy efficiency.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.