A Quick And Generic Approach Of Selecting Compiler Optimization Options

As the popularity of computing technology grows, there is higher and higher demand on high performance computation. As a consequence, software optimization, an effective approach to boost performance, is getting more and more important. The term of software optimization covers a number of optimization algorithms. However, in real-life projects, engineers make use of software optimization by selecting compiler options, so, in most cases, selecting compiler options means selecting optimization algorithms. The performance of target binaries, which are generated with different option combinations, differs greatly, because of application behavior, target platform, data layout, code layout and interactions between optimizations. So it's important to select the best compiler options for every application. But unfortunately, selecting compiler options is not straightforward. Without a methodology, option selection is usually blindly or empirical. So, a general and quick methodology of options selection is valuable.There hasn't been any report on domestic research work on compiler options selection, while the research work going on in foreign countries is limied in that:1. They're not general because they focus on specific target platform.2. They're not doable in most real-life projects because they require modifying compiler itself.Based on related works and engineers'knowledge, this paper proposes a group-based methodology of compiler option selection. The methodology is featured with effectiveness, general and automatic. It's especially useful in developing high performance computation or media processing programs.