A high-performance mapping algorithm for heterogeneous computing systems

A mapping algorithm for heterogeneous computing systems is proposed in this paper. This algorithm utilizes a new indicator-the relative cost-to obtain optimal mapping. The existing Min-min algorithm can be well explained under synergy of this new indicator. It is found that the Min-min algorithm leaves room for improvement because of its haste to reduce completion time by overlooking the impact of load balance. Our new algorithm retains the advantages of the Min-min algorithm and balances the load very well. It demonstrates the ability to generate good mapping in various heterogeneous environments.