1732 IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, VOL. 33, NO. 11, NOVEMBER 2014 Design Space Exploration for Wireless NoCs Incorporating Irregular Network Routing Paul Wettin, Member, IEEE, Ryan Kim, Student Member, IEEE, Jacob Murray, Member, IEEE, Xinmin Yu, Student Member, IEEE, Partha P. Pande, Senior Member, IEEE, Amlan Ganguly, Member, IEEE, and Deukhyoun Heo, Senior Member, IEEE Abstract—The millimeter-wave small-world wireless network- on-chip (mSWNoC) is an enabling interconnect architecture to design high-performance and low-power multicore chips. As the mSWNoC has an overall irregular topology, it is essential to design and optimize suitable deadlock-free routing mechanisms for it. In this paper, we quantify the latency, energy dissipation, and thermal profiles of mSWNoC architectures by incorporat- ing irregular network routing strategies. We demonstrate that the latency, energy dissipation, and thermal profile are affected by the adopted routing methodologies. The overall system perfor- mance and thermal profile are governed by the traffic-dependent optimization of the routing methods. Our aim is to establish the energy-thermal-performance trade-offs for the mSWNoC depending on the exact routing strategy and the characteristics of the benchmarks considered. Index Terms—Irregular networks, millimeter-wave wireless, network-on-chip (NoC), routing algorithms, small-world. I. I NTRODUCTION W IRELESS network-on-chip (WiNoC) is envisioned as an enabling technology to design low-power and high- bandwidth, massive multicore architectures [1]. The existing method of implementing a NoC with planar metal interconnects is deficient due to high latency, significant power consumption, and temperature hotspots arising out of long, multihop wireline paths used in data exchange. It is possible to design high- performance, robust, and energy-efficient multicore chips by Manuscript received December 11, 2013; revised May 27, 2014; accepted August 1, 2014. Date of current version October 16, 2014. This work was supported in part by the U.S. National Science Foundation (NSF) under Grant CCF-0845504, Grant CNS-1059289, and Grant CCF-1162123, and in part by the Army Research Office under Grant W911NF-12-1-0373. This paper was recommended by Associate Editor L. P. Carloni. P. Wettin is a Senior ASIC Design Engineer with Marvell Semiconductor, Boise, ID, USA. He did this work as a PhD student with the School of Electrical Engineering and Computer Science, Washington State University, Pullman, WA 99164 USA (email: pwettin@eecs.wsu.edu). J. Murray is a Clinical Assistant Professor with the Department of Electrical Engineering and Computer Science, Washington Statue University, Everett, WA, USA. He did this work as a PhD student with the School of Electrical Engineering and Computer Science, Washington State University, Pullman, WA 99164 USA (email: jmurray@eecs.wsu.edu). R. Kim, X. Yu, P. P. Pande, and D. Heo are with the School of Electrical Engineering and Computer Science, Washington State University, Pullman, WA 99164 USA (email: rkim@eecs.wsu.edu; xyu@eecs.wsu.edu; pande@eecs.wsu.edu; dheo@eecs.wsu.edu). A. Ganguly is with the Department of Computer Engineering, Rochester Institute of Technology, Rochester, NY 14623 USA (e-mail: amlan.ganguly@rit.edu). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TCAD.2014.2351577 adopting novel architectures inspired by complex network the- ory in conjunction with on-chip wireless links. Networks with the small-world property have very short average path lengths, making them particularly interesting for efficient communica- tion with minimal resources. Using the small-world approach we can build a highly efficient NoC with both wired and wire- less links. Neighboring cores should be connected through traditional metal wires while widely separated cores will com- municate through long-range, single-hop, wireless links. A small-world network principally has an irregular topology [2]. Routing in irregular networks is more complex, because routing methods are typically topology agnostic. Hence, it is necessary to investigate suitable routing mechanisms for small-world net- works. Routing in irregular networks can be classified into two broad categories, viz., rule- and path-driven strategies [3]. Rule-driven routing is typically done by employing a span- ning tree for the network. Messages are routed along this spanning tree with specific restrictions to achieve deadlock freedom. Because deadlock freedom is taken into account first for these routing strategies, minimal paths through the net- work for every source-destination pair cannot be guaranteed [3]. Conversely, for path-driven routing, minimal paths between all source-destination pairs are first guaranteed and then deadlock freedom is achieved by restricting portions of traffic from using specific resources such as the virtual channels [3]. We follow the above-mentioned strategies to design suitable routing mechanisms for the millimeter (mm)-wave small- world wireless NoC (mSWNoC) [4]. In the rule-based routing, a spanning tree of the network is created where data is routed along the spanning tree. An allowed route never uses a link in the up direction along the tree after it has been in the down direction once. Hence, channel dependency cycles are pro- hibited, and deadlock freedom is achieved [5]. However, a well-known weakness of this routing scheme is that it has a strong tendency to generate hotspots around the root of the tree structure. In the path-based routing, the network resources are divided into layers and network deadlocks are avoided by pre- venting portions of traffic from using specific layers [3], [6]. The achievable performance of mSWNoC depends on the effi- ciency of these routing algorithms. The power and thermal profiles of the system depend on how efficiently the routing mechanisms can move the traffic through the network while balancing the traffic among the network elements. These routing algorithms have previously been studied for traditional parallel computing systems where the comparative 0278-0070 c  2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.