병렬 컴퓨터 시스템에서의 FFT 데이터 흐름도에 관한 유전 스케줄링 알고리즘

Genetic Scheduling Algorithm for FFT Dta Flows in Parallel Computers

We propose the genetic algorithm to apply three kinds of FFT data flows to be considered the overhead for the data exchange between processors that have the multi-scheduling problem on parallel computer In the design of genetic algorithm, we propose the chromosome representation which can simply encode and decode a solution without any heuristic information, the evaluation function to be considered an efficiency of processor, and the genetic operator to inherit a superior gene from their parents. And we saw that the simulation result can verify better performance than the existing algorithm(BEA : binary exchange algorithm)in the face of execution time.