Bilkent University
Department of Computer Engineering

---

S E M I N A R

 

Decomposition of Workload Arrays for Heteregenous Systems

 

Ertugrul Kartal Tabak
Ph.D Student
Computer Engineering
Bilkent University

The decomposition of workload arrays to processors is an important task in parallel and distributed computing. The one-dimensional decomposition of workload arrays to an array of processors can be modeled by chain-on-chain problem. There is a rich literature for the homogenous version of the problem, where the processing power of processors are assumed to be equal. Heuristics for heteregonous chain-on-chain problems are still applied with the hope to ease implementation and achieve reasonable decompositions. We have studied on the exact solutions of heteregonous CCP problem, where the order of processors is specified. We have revised the algorithms for homogenous systems. The tests proved that heuristics are overbeaten by the exact solutions for heteregonous systems.

 

DATE: May 1, 2006, Monday@ 16:40
PLACE: EA 409