To find a good set of options for a particular CPU and particular software is actually a difficult task. Some tool based on genetic algorithm like AcovEA exists, it will compile and run each program a lot of times, this is a obviously a huge problem if we want to use this tool with programs who have a long execution time, to try bigger problem, the classical example could be to optimize a long genetic algorithm search. The objective of this project is to develop a tool (we name it AcovSA) with the same purpose as AcovEA but based on Simulated Annealing. With bigger benchmark given – scheduling problem with some test data DAG (Directed Acyclic Graph) and finally find the good OOS (optimization options set), from more than 60 optimization options, for compiling the programs with GCC C compiler on a Linux platform. At last, comparing the benchmarks’ running time by compiling with the good OOS obtain from AcovSA and AcovEA to see both the advantages and disadvantages on two methods.