CAREER: Application-Driven Approach for Building Run-Time Systems for Future Teraflop and Petaflop Architectures

点击翻译按钮获取中文摘要

CCR-9876179"CAREER: Application-driven Approach for Prototyping Run-Time Systems for Future Teraflops and Petaflops Architectures"Nikos P. ChrisochoidesRecent feasibility studies for the next generation of Teraflops andPetaflops parallel architectures indicate that these machines will bemulti-layer systems that consist of many layers of similar componentsfor memory, processors, and interconnects with different levels ofperformance. Such designs will widen the latency gap between memoryand CPU access to a point where no form of CPU-initiated data movementfrom memory will be satisfactory. To solve this, research needs to bedirected towards hardware like "smart" memories that are pro-activewith regard to memory-transfers, algorithms, and software. Thisresearch focuses on the development and evaluation of latency tolerantalgorithms and run-time software support systems for adaptivecomputations on a multi-layer parallel platform that is "isomorphic"to future Teraflops and Petaflops systems. This project will have animmediate impact on Petaflops Hybrid Technology Multi-Threaded (HTMT)design by providing hardware designers with feedback based on thebehavioral and structural analysis of an end-to-end unstructured andadaptive application from fracture mechanics. Also, this project willprovide a testbed for training the next generation of computationalscientist to effectively use such multi-layer parallel platforms andhelp computer engineers to understand the hardware requirements oflarge-scale applications.