A NOVEL APPROACH FOR SELECTION OF BEST SET OF OPTIMIZATION FUNCTIONS FOR A BENCHMARK APPLICATION USING AN EFFECTIVE STRATEGY

Finding right set of optimization techniques for a given application involves lot of complications. The compiler optimization technique for a given platform depends upon the various factors such as hardwaresettings and problem domain as well as orderings. Recent version of GCC compiler consists of more number of optimization techniques. Applying all these techniques to a given application is not feasible,because of program performance degradation. Searching best set of optimal techniques as well as orderings for an application is an extremely critical and challenging task. Many previous works tries toreduce the search space, but such approaches take more time and also expensive. Previously machine learning algorithm has been used to predict best set of sequences, but it requires longer training phaseand more data sets. In this paper we have proposed an efficient orchestration algorithm such as optimality random search and advanced combined elimination, which selects optimal set from more than100 techniques. Result shows that advanced combined elimination works well for most of the benchmark applications than optimality random search.