Benchmarking and performance enhancement framework for multi-staging object-oriented languages

This paper focuses on verifying the readiness, feasibility, generality and usefulness of multi-staging programming in software applications. We present a benchmark designed to evaluate the performance gain of different multi-staging programming (MSP) languages implementations of object oriented languages. The benchmarks in this suite cover different tests that range from classic simple examples (like matrix algebra) to advanced examples (like encryption and image processing). The benchmark is applied to compare the performance gain of two different MSP implementations (Mint and Metaphor) that are built on object oriented languages (Java and C# respectively). The results concerning the application of this benchmark on these languages are presented and analysed. The measurement technique used in benchmarking leads to the development of a language independent performance enhancement framework that allows the programmer to select which code segments need staging. The framework also enables the programmer to verify the effectiveness of staging on the application performance. The framework is applied to a real case study. The case study results showed the effectiveness of the framework to achieve significant performance enhancement.