from    
to    
search  

 


Protein Mechanics: from Single Molecule Force Spectroscopy toProtein-based Bi...
化工系膜中心学术论坛-MOF Chemistry: From design strategies to Applications
清华大学材料科学与工程研究院《材料科学论坛》学术报告:Multi-aspect characteri...
Brain-like spiking neural networks: A 4th generation of neural network models
报告题目:
Architecture and Algorithm Design to Improve the Concurrency of Multi-core Computing
 报告人:
Bo Hong
Dr. 
Assistant Professor
ECE Department of Drexel University
报告时间:
2008-06-23 15:00
报告地点:
3-125, FIT Building
主办单位:
信研院
  简介:

2008年清华大学信息技术研究院系列学术报告10

 Dr. Bo Hong is currently an assistant professor at the ECE Department of Drexel University. He received his Ph.D. degree in Computer Engineering from the University of Southern California in 2005. Prior to that, he received his Master and Bachelor's degree from Tsinghua University, China in 2000 and 1997, respectively.

Dr. Bo Hong's research interests include High performance computing, multi-core computer architecture, parallel and multi-threaded algorithms, and distributed computing. His current research effort is concentrating on thread synchronization in multi-core processors. In this hardware-software integrated research, new algorithms are being explored to improve the efficiency of multi-core computing through new synchronization circuitry.

 

Abstract

Given the increasing emphasis on multi-core architectures, the extent to which an application can be multi-threaded to keep the multiple processor cores busy is likely to be one of the greatest constraints on the performance of next generation computing platforms. However, Amdahl’s law has pointed out that only a portion of a program’s processing can be run in parallel. Only this portion scales with machines having more and more processors. The rest of the work load is sequential and needs to be protected from shared accesses, and locking is the mostly used method. With the current industry trend where almost all hardware manufacturers are now aggressively exploring multi-core platforms, lock contention is becoming a pressing concern for developers looking to write performance code. In this talk, the design and performance drawbacks of locks will be examined, together with recent development in the alternatives of locking. As an illustrative example, a new lock-free multi-threaded algorithm for the maximum network flow problem will be presented with significantly improved execution speed and scalability on multi-core architectures.
今日相关信息
Discrete Late Metal Nitrenes in Catal...
低碳能源实验室学术讲座:Geological St...
 
同类别相关信息
The Design and Implementation of an...
清华论坛第78讲:全球气候治理与中美气...
清华大学国际纳米光电子学研究中心成立仪...
量子计算:从大数分解到大数据处理
量子计算: 一场从根上开始的革命
学术活动