邮箱登录 | 所务办公 | 收藏本站 | English | 申博sunbet手机版登陆科学院
 
首页 申博sunbet中文版概况 新闻动态 科研成果 研究队伍 国际交流 技术转移 研究生教育 学术出版物 党群园地 科学传播 信息公开
国际交流
学术活动
交流动态
现在位置:首页 > 国际交流 > 学术活动
High-Performance Software Development Challenges in the Post-Moore Era
2019-07-26 | 【 【打印】【关闭】

  Abstract: The end of Moore's Law scaling for VLSI technology implies that significant performance increases for future generations of processors cannot derive from increased transistors counts. Instead, hardware customization and more efficient use of hardware resources are expected to be primary means of performance improvement. Hence, the already challenging task of application software development will get even harder. Advances in software infrastructure such as compilers will be crucial to assist application developers achieve high-performance without loss of productivity and portability.

  A very fundamental challenge faced by compilers is data-locality optimization. The cost of data movement far exceeds the cost of performing arithmetic/logic operations on current processors, both in terms of energy as well as execution time. But while the computational complexity of most practically used algorithms is quite well understood, the same is not true of data-movement complexity. There is a need to develop new abstractions and methodologies, and create tools for characterization and optimization of data movement. This talk will discuss challenges and some promising directions in the quest to achieve the three desirables of performance, productivity, and portability in the development of high-performance software.

  Bio:

  Sadayappan is a Professor in the School of Computing at the University of Utah, with a joint appointment at Pacific Northwest National Laboratory. He was previously a Professor of Computer Science and Engineering and a University Distinguished Scholar at the Ohio State University. His primary research interests center around performance optimization and compiler/runtime systems for high-performance computing, with a special emphasis on high-performance frameworks that enable high productivity for application developers. He collaborates closely with computational scientists and data scientists in developing high-performance domain-specific frameworks and applications. Sadayappan received a B.Tech from the Indian Institute of Technology, Madras, and M.Sc. and Ph.D. from Stony Brook University, all in Electrical Engineering. Sadayappan is an IEEE Fellow.

 
网站地图 | 联系我们 | 意见反馈 | 所长信箱
 
京ICP备05002829号 京公网安备1101080060号