Heterogeneous Parallel Embedded Systems

Embedded systems are driving an informationrevolution with their pervasion in our everyday lives. With more and more different types of FUsavailable, same type of operations can be processedby heterogeneous FUs with different costs, where thecost may relate to power, reliability, etc. Due toconditional operations, some tasks may not have fixedexecution time. Therefore, how to assign a proper FUtype to each operation of a system and generate aschedule that minimize the total costs whilesatisfying timing constraints with guaranteedconfidence probabilities becomes a critical issue.This book, 1) proposed several efficient algorithmsto solve the problem; 2) combined Dynamic VoltageScaling (DVS) and soft real-time to solve the VoltageAssignment with Probability (VAP) Problem. 3) appliedour efficient algorithms to dynamic adjust theworking mode of sensors to save energy for sensornetworks; 4) designed rotation scheduling algorithmsfor real-time applications that produce schedulesconsuming minimal energy; 5) combined data mining andprefetching to reduce energy consumptions; 6)improved performance and energy consumption forprocessors with multi-bank memory.

Edwin Shareceived Ph.D. of CS from Princeton University. Since 2000, hehas been a tenured full professor in the Dept. of CS at UTD andpublished more than 220 papers.