STTR Phase I: Visualization API Enablers for a High-End Fine-Grained Parallel Processor

STTR 第一阶段：高端细粒度并行处理器的可视化 API 启用器

• 批准号: 0339489
• 负责人: Uzi Vishkin
• 金额: $ 10万
• 依托单位: XMTT, Inc.
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2004
• 资助国家: 美国
• 起止时间: 2004-01-01 至 2004-12-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0339489&HistoricalAwards=false
• 关键词: STTR; Phase; Visualization; API; Enablers

This Small Business Technology Transfer Phase 1 project seeks to advance a major application domain for a recent patented technology, Explicit-Multi-Threading (XMT), for building and using computers. Numerous algorithm researchers have developed a computational model, the Parallel Random Access Model (PRAM), during the 1980s and 1990s. From that starting point, a highly parallel XMT processor architecture whose performance objective is reducing single task completion time has been conceived and developed. Using fast driving as a metaphor for fast single task computing, and a flat tire for cache miss, XMT seeks not only to build cars that can be driven very fast; the cars will also allow safely changing a flat tire while the fast driving continues.The XMT technology will give programmers the freedom and power to think and code in parallel. The simplicity of the parallel programming model enables high productivity for the application programmer. Orders of magnitude faster general purpose single task completion time for a given amount of hardware are expected.

这个小企业技术转让第一阶段项目旨在推进一个主要的应用领域，为最近的专利技术，显式多线程（XMT），为建设和使用计算机。在20世纪80年代和90年代，许多算法研究人员开发了一种计算模型，即并行随机存取模型（PRAM）。从这个出发点，一个高度并行的XMT处理器架构，其性能目标是减少单个任务完成时间已经构思和开发。使用快速驾驶作为快速单任务计算的隐喻，以及高速缓存未命中的爆胎，XMT不仅寻求建造可以非常快速地驾驶的汽车;汽车还允许在快速驾驶的同时安全地更换爆胎。XMT技术将给予程序员并行思考和编码的自由和力量。并行编程模型的简单性使应用程序员能够实现高生产力。对于给定数量的硬件，通用单个任务的完成时间有望提高几个数量级。