CAREER: Compiler and Run-time System Support for Modern Hardware

要在现代计算机上实现最佳性能，必须了解和利用程序行为、运行时系统和硬件之间的交互。本研究探索编译器和运行时系统对提高负载值预测和降低程序能耗的支持。加载值预测通过允许加载与依赖它的指令并行执行来暴露程序中的指令级并行性。能量减少技术在不显著降低程序性能的情况下降低了程序的能量需求。本研究主要考虑两种技术：(I)使用程序中容易获得的信息(如类型信息)来转换或注释代码以更好地利用硬件的编译器技术；(Ii)通过重新组织和注释数据以更好地利用硬件的运行时系统技术。这项研究还将考虑可在软件中利用的新的负荷值预测和能源消耗的硬件机制。

To achieve the best performance on modern computers, one mustunderstand and exploit the interactions between program behavior,run-time systems, and hardware. This research explores compiler andrun-time system support for improving load-value prediction and forreducing the energy consumption of programs. Load-value predictionexposes instruction-level parallelism in programs by allowing loads toexecute in parallel with the instructions that depend on it. Energyreduction techniques reduce the energy requirements of programswithout significantly degrading performance.Broadly speaking this research considers two kinds of techniques: (i)Compiler techniques that use easily-available information in programs(such as type information) to transform or annotate code to makebetter use of hardware; (ii) Run-time system techniques thatreorganize and annotate data to make better use of hardware. Thisresearch will also consider new hardware mechanisms for load-valueprediction and energy consumption that can be exploited in software.