CAREER: Adaptive Concurrency Management for Multicore Computing

Georgia Tech Research Corporation, Atlanta GA

Investigators

Abstract

Career: Adaptive Concurrency Management for Multicore Computing Abstract: Given the increasing emphasis on multicore computing, concurrency management is likely to be one of the key techniques to unleash the power of multicore processors. Concurrency is the capability of executing multiple tasks simultaneously, but it is often difficult to achieve due to data and control dependencies. The proposed research aims to optimize thread-level concurrency for multicore applications. It will investigate architecture features for efficient shared data accesses, and use these features to dynamically control the advancement of concurrent threads and the allocation of CPU resources. The project will target computing platforms of immediate concern and the techniques will be made accessible to the programming community by integrating them into existing programming tools. The project will also extend theoretical results towards efficient multi-threaded algorithms. The proposed research project is expected to significantly improve the performance of multicore computing and expand the range of applications that can benefit from such processor platforms. It is expected to facilitate efficient multicore processing for computation-demanding applications in science, engineering, and business.