SHF: CSR: Medium: Collaborative Research: Hierarchical On-Chip Millimeter-Wave Wireless Micro-Networks for Multi-Core Systems

带有数百个嵌入式功能块的多核处理器旨在为图形、金融和科学建模、网络、多媒体和无线基础设施等各种应用程序实现前所未有的性能优势。片上网络(NoC)是一种能够在单个芯片上集成多个嵌入式内核的方法。然而，随着集成度的不断提高，由于传统的基于金属/介质的互连，传统的片上网络面临着片上数据传输的高延迟和能量消耗问题。无线NoC(WiNoC)通过将多跳有线链路替换为高带宽单跳远程无线信道，同时解决了传统NoC的延迟、功耗和互连路由问题。最近的研究已经表征了工作在50-110 GHz毫米波范围内的硅集成片上天线，这现在是一种可行的技术。再加上毫米波收发器设计的重大进步，这一进展为设计高带宽和低功耗的WiNoC开辟了新的研究机会。拟议的研究将为片上无线微网络开发跨层设计方法。将为WiNoC开发合适的物理层、数据链路层和网络层协议。这项工作的预期贡献是宽带和可积毫米波WiNoCs的开发和演示。这一提议的变革性方面在于从一个全新的角度解决芯片内互连问题，即通过开发微米/纳米级的宽带毫米波无线网络，而不是追求传统有线互连的渐进式改进。通过将无线网络引入片上环境，拟议的研究将在多核芯片设计中引入一种范式转变。拟议的研究将促进本科生和研究生的教育，使他们能够将课堂知识应用于需要硬件、软件和理论专业知识的研究问题。这种专业知识将在大学和工业实验室环境中使用最先进的测试设备和计算设施来开发，最终将是保持美国技术领先地位所必需的。来自不同代表性不足群体的学生，包括妇女、非裔美国人和拉美裔美国人，将参与这一项目。该项目的教育影响将得到预期研究成果将对整个社会产生的积极影响的补充。

Multi-core processors with hundreds of embedded functional blocks are designed to achieve unprecedented performance benefits for a wide variety of applications such as graphics, financial and scientific modeling, networking, multimedia and wireless infrastructure. The Network-on-Chip (NoC) is an enabling methodology to integrate many embedded cores on a single die. However, with growing levels of integration traditional NoCs suffer from high latency and energy dissipation in on-chip data transfer due to conventional metal/dielectric based interconnects. The wireless NoC (WiNoC) simultaneously addresses the latency, power consumption and interconnect routing problems of conventional NoCs by replacing multi-hop wired links with high-bandwidth single-hop long-range wireless channels. Recent investigations have characterized silicon integrated on-chip antennas operating in the millimeter (mm)-wave range of 50-110 GHz, and this is now a viable technology. Coupled with significant advances in mm-wave transceiver design, this development opens up new research opportunities for designing high bandwidth and low power WiNoCs.The proposed research will develop a cross-layer design methodology for an on-chip wireless micro-network. Suitable physical, data link and network layer protocols will be developed for the WiNoC. The expected contribution of this work is the development and demonstration of broadband and integrable mm-wave WiNoCs. The transformational aspect of this proposal lies in addressing the on-chip interconnect problem from a fundamentally new perspective, namely by developing a broadband millimeter wave wireless network at the micro/nanoscale, as opposed to pursuing incremental improvements in traditional wired interconnects. By bringing wireless networking to the on-chip environment, the proposed research will introduce a paradigm shift in the design of multi-core chips. The proposed research will facilitate the education of undergraduate and graduate students by allowing them to apply classroom knowledge to a research problem requiring hardware, software and theoretical expertise. Such expertise will be developed in both university and industrial laboratory settings using state-of-the-art test equipment and computing facilities, and will ultimately be necessary to maintain the technological leadership of the United States. Students from various underrepresented groups, including women, African Americans and Hispanics, will be engaged in this project. The project's educational impacts will be complemented by the positive effect that the research outcomes are expected to have on society as a whole.