SHF: Large: Rethinking the Architecture of FPGAs as First-Class Computing Devices

SHF：大型：重新思考 FPGA 作为一流计算设备的架构

• 批准号: 1012851
• 负责人: James Hoe
• 金额: $ 100万
• 依托单位: Carnegie-Mellon University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2010
• 资助国家: 美国
• 起止时间: 2010-09-01 至 2015-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1012851&HistoricalAwards=false
• 关键词: SHF; Large; Rethinking; Architecture; FPGAs

The CORAM (Connected-RAM) project endeavors to completely rethink the reconfigurable computing fabric from scratch to create a new generation of FPGAs as first-class computing devices. From a computing perspective, a major deficiency in today?s FPGAs is the lack of a native memory architecture. The goal of the CORAM project is to develop a consistent and hardware-friendly memory architecture for FPGAs. The CORAM project will develop (1) convenient memory interface abstractions to support the application developers in accessing on- and off-chip memory and (2) the associated mechanisms, from the architecture-level down to the circuit-level, to efficiently support those memory abstractions. The CORAM project will study the abstractions and mechanisms using a combination of FPGA-based prototypes and a test-chip prototype.The CORAM project is motivated by the renewed interest in FPGA-based computing, driven by the need for greater computing performance and at the same time greater energy efficiency. Reconfigurable computing is one of the key technology candidates to extend the exponential performance scaling efficiently beyond the current commercial multicore processor technology horizon. The CORAM project is central to fully exploring the potential of the reconfigurable computing paradigm.

CORAM（Connected-RAM）项目致力于从头开始重新思考可重构计算结构，以创建新一代FPGA作为一流的计算设备。从计算的角度来看，今天的一个主要缺陷是什么？的FPGA是缺乏一个本机存储器架构。 CORAM项目的目标是为FPGA开发一种一致的、硬件友好的存储器架构。 CORAM项目将开发（1）方便的存储器接口抽象，以支持应用程序开发人员访问片上和片外存储器，以及（2）从架构级到电路级的相关机制，以有效地支持这些存储器抽象。 CORAM项目将使用基于FPGA的原型和测试芯片原型的组合来研究抽象和机制。CORAM项目的动机是对基于FPGA的计算的重新兴趣，以及对更高计算性能和同时更高能效的需求。 可重构计算是将指数级性能扩展有效地扩展到当前商业多核处理器技术水平之外的关键技术候选之一。 CORAM项目是充分探索可重构计算范式潜力的核心。