Principle Investigators:
This project concerns a number of different issues in processor scheduling, synchronization, and workload allocation in parallel systems, as well as the interplay between such issues and scalability issues as processors become relatively faster compared to memories and interconnects. Also of interest are the design and performance evaluation of various flow-control and routing strategies for multiprocessor interconnection networks employing wormhole switching (a technique in which messages are pipelined through the network).
One of the longer-term goals of the project has concerned run-time support systems for parallel computing. Of particular interest are systems that relieve the programmer and compiler from concerns of granularity, scheduling, and the enforcement of synchronization constraints, through efficient and flexible run-time mechanisms. Much work has been accomplished in the design and performance evaluation of such mechanisms for small-scale shared-memory multiprocessors. Of current interest are high-latency environments such as networks of workstations.
 Research
Areas |
 DISCUS
|
 Research
Groups |
 Comp
Sci. Dept |
 University |