CRI: IAD Resources for Branch Prediction Research

拟议的基础设施购置和开发将提供计算资源，以加强PI部门的教职员工和学生进行计算即时研究项目的能力。一个由32台计算机组成的集群将为在分支预测领域进行的编译器和计算机体系结构研究所产生的任务提供高吞吐量。准确的分支预测使处理器能够预测和响应未来事件，以便提高性能，同时将错误路径推测导致的能耗保持在较低水平。该基础设施将允许测量在多核处理器上运行的程序的性能和能量，并允许对编译器、处理器微体系结构和各种分支预测器方案进行建模，以最好地响应程序执行中发生的变化。

The proposed infrastructure acquisition and development will provide computing resources to enhance the ability of faculty and students in the PI's department to do computationally instensive research projects. A cluster of 32 computers will provide high throughput for tasks arising from compiler and computer architecture research done on in the area of branch prediction. Accurate branch prediction enables a processor to anticipate and respond to future events in order to improve performance while keeping energy consumption due to wrong-path speculation low. The infrastructure will allow the performance and energy of programs running on multicore processors to be measured and enable the modeling of compiler, processor microarchitecture and various branch predictor schemes which can best respond to changes occuring in program execution.