Abstract:

High level program transformations are critical to improve applications’ performance,many of which mainly concern the determination of optimal transformation parameters, such as loop blocking.Since optimization parameter search problem is featured NP-hard,to date,no one can find any deterministic algorithms to efficiently solve this problem.Considering it as a nonlinear global optimization problem,we introduce an improved simulated annealing algorithm to find the optimal parameters.Several comparative experiments demonstrate the performance and effectiveness of the new method.

Key words: performance optimization of programs;optimization transformation;optimization parameters;simulated annealing