Unconventional fabrics, architectures, and models for future multi-core systems

R. Marculescu, C. Teuscher, P. Pande
{"title":"Unconventional fabrics, architectures, and models for future multi-core systems","authors":"R. Marculescu, C. Teuscher, P. Pande","doi":"10.1145/1878961.1879017","DOIUrl":null,"url":null,"abstract":"Massive level of integration is making modern multi-core chips all-pervasive in several domains. Hence, high performance, robustness, and low power are crucial for the widespread adoption of such platforms. However, achieving all these goals forces us to re-think the basis of designing multi-core systems at nanoscale, starting with the very substrate we need to use to implement such systems in the future, particularly for nanowire (or carbon nanotube) based on-chip interconnect obtained through self-assembly techniques. Due to the lack of control over these processes, such interconnects are expected to be largely unstructured. While large unstructured networks are easy to fabricate, they require unconventional architectures and communication paradigms. For instance, by getting inspiration from many natural systems with network-based architectures, the future multi-core systems at nanoscale are expected to be hierarchical and heterogeneous in nature, as many powerful features such as increased performance, better resource utilization, and an increased robustness against failures of many natural networks come precisely from their heterogeneity, unstructuredness, and hierarchical nature. As such, an important performance limitation of multi-core chips designed with regular network architectures arises from planar metal interconnect-based multi-hop links, where the data transfer between two distant blocks can cause high latency and power consumption.","PeriodicalId":118816,"journal":{"name":"2010 IEEE/ACM/IFIP International Conference on Hardware/Software Codesign and System Synthesis (CODES+ISSS)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2010-10-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2010 IEEE/ACM/IFIP International Conference on Hardware/Software Codesign and System Synthesis (CODES+ISSS)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1145/1878961.1879017","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0

Abstract

Massive level of integration is making modern multi-core chips all-pervasive in several domains. Hence, high performance, robustness, and low power are crucial for the widespread adoption of such platforms. However, achieving all these goals forces us to re-think the basis of designing multi-core systems at nanoscale, starting with the very substrate we need to use to implement such systems in the future, particularly for nanowire (or carbon nanotube) based on-chip interconnect obtained through self-assembly techniques. Due to the lack of control over these processes, such interconnects are expected to be largely unstructured. While large unstructured networks are easy to fabricate, they require unconventional architectures and communication paradigms. For instance, by getting inspiration from many natural systems with network-based architectures, the future multi-core systems at nanoscale are expected to be hierarchical and heterogeneous in nature, as many powerful features such as increased performance, better resource utilization, and an increased robustness against failures of many natural networks come precisely from their heterogeneity, unstructuredness, and hierarchical nature. As such, an important performance limitation of multi-core chips designed with regular network architectures arises from planar metal interconnect-based multi-hop links, where the data transfer between two distant blocks can cause high latency and power consumption.
未来多核系统的非常规结构、架构和模型
大规模集成度使得现代多核芯片在多个领域无处不在。因此,高性能、健壮性和低功耗对于这种平台的广泛采用至关重要。然而,实现所有这些目标迫使我们重新思考在纳米尺度上设计多核系统的基础,从我们未来实现这些系统所需的衬底开始,特别是通过自组装技术获得的基于片上互连的纳米线(或碳纳米管)。由于缺乏对这些过程的控制,这种互连预计将在很大程度上是非结构化的。虽然大型非结构化网络很容易构建,但它们需要非常规的架构和通信范式。例如,通过从许多基于网络架构的自然系统中获得灵感,未来的纳米级多核系统有望在本质上是分层的和异构的,因为许多强大的特性,如提高性能、更好的资源利用率和增强对许多自然网络故障的鲁棒性,正是来自于它们的异构性、非结构化和分层性质。因此,使用常规网络架构设计的多核芯片的一个重要性能限制来自基于平面金属互连的多跳链路,其中两个遥远块之间的数据传输可能导致高延迟和功耗。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信