ITR: Dynamic Cooperative Performance Optimization

面向对象的编程语言，特别是Java，因为它们的好处而得到广泛使用，这些好处很大程度上来自于它们的灵活性。但同样的灵活性使得Java程序更难提前优化。一个预先优化的程序样本集合将40%-95%的时间花费在积极的处理器上，等待内存提供数据，这说明需要改善内存行为。未来的处理器只会使问题变得更糟。该项目对这个问题进行了综合攻击，包括新的程序分析和优化，配置文件反馈，运行时技术，包括自适应垃圾收集算法，以及将高级预测和程序行为观察传达到硬件的方法。该项目的目标是设计和构建一个编译器，运行时系统，以及增强的体系结构和操作系统功能，以快速和优雅地对编译器预测和实际运行时行为做出反应，以实现高性能。该项目的目标是通过系统组件之间的协作来实现协同，而不是在单个组件内解决每个问题。虽然研究的重点是提高内存性能，但该项目的感知框架适用于广泛的性能优化技术，因此预期的研究成果和软件产品具有更广泛的影响。

Object-oriented programming languages, notably Java, are gaining broad usebecause of their benefits, which come largely from their flexibility. Butthis same flexibility makes Java programs more difficult to optimize inadvance. One sample collection of programs, optimized in advance, spent40-95% of their time on an aggressive processor waiting for the memory toprovide data, illustrating the need to improve memory behavior. Futureprocessors will only make the problem worse.The project makes an integrated attack on this problem, incorporating newprogram analyses and optimizations, profile feedback, run-time techniquesincluding adaptive garbage collection algorithms, and ways of communicatinghigh-level predictions and observations of program behavior to thehardware.The project aims to design and build a compiler, run-time system, andenhanced architectural and operating system features that react quickly andgracefully to compiler predictions and actual run-time behavior to achievehigh performance. The goal is synergy via cooperation between the systemcomponents versus solving each problem within an individual component.While the research focuses on improving memory performance, the project'senvisioned framework is suited to a wide range of performance optimizationtechniques, so the expected research results and software products havebroader impact.