As process technologies have scaled, the increasing number of processor cores and memories on a single die has also driven the need for more complex on-chip interconnection networks. Crossbar switches are primary building blocks in such networks-on-chip, as they can be used as fast single-stage networks or as the core of the router switch in multi-stage networks. While crossbars offer non-blocking, single-hop, all-to-all communication, they tend to scale poorly with the number of nodes due to the latency and energy of the long wires and high-radix multiplexor structures needed. To combat these limitations, we propose a low-swing crossbar design that uses capacitively driven wires and capacitively coupled multiplexers. Capacitively driven wires offer low swing signaling, higher bandwidths, and low energy consumption, while capacitively coupled multiplexers offer reduced parasitic loading from the inactive inputs. We present a 16×16 64b low-swing crossbar switch designed in a TSMC 40nm CMOS bulk process. Post-layout simulation shows it operating at a maximum frequency of 2.2GHz, achieving a bandwidth of 2.56Tb/s at 0.9V (nominal Vdd) with an area of 0.94mm2. Total energy consumption for full, half, and minimum bandwidths are 110pJ, 84pJ, and 64pJ respectively, thus offering an efficiency of 10.49 Tbps/W, a 3X improvement over previously published results.

The crossbar is a popular topology for on-chip networks that offers non-blocking connectivity and uniform latency. However, as the number of nodes increases, crossbars typically scale poorly in area, power, and latency/throughput. To better understand the design space, we have developed an on-chip crossbar modeling tool based on analytical models calibrated using circuit-level simulation results in 40nm CMOS. We present a design space exploration showing how crossbar area, power, and performance vary across input/output node number, data width, wire parameters, and circuit implementation. Using the modeling results, we identify a design point that demonstrates 2X higher throughput, 1.4X lower power and 1.2X lower area compared to previous published designs.

Silicon photonics is a promising technology to scale offchip bandwidth in a power-efficient manner. Given equivalent bandwidth, the flexibility of switched networks often leads to the assumption that they deliver greater performance than point-to-point networks on message passing applications with low-radix traffic patterns. However, when optical losses are considered and total optical power is constrained, this assumption no longer holds. In this paper we present a power constrained method for designing photonic interconnects that uses the power characteristics and limits of optical switches, waveguide crossings, inter-layer couplers and waveguides. We apply this method to design three switched network topologies for a multi-chip system. Using synthetic and HPC benchmark-derived message patterns, we simulated the three switched networks and a WDM point-to-point network. We show that switched networks outperform point-to-point networks only when the optical losses of switches and inter-layer couplers losses are each 0.75 dB or lower; achieving this would require a major breakthrough in device development. We then show that this result extends to any switched network with similarly complex topology, through simulations of an idealized "perfect" network that supports 90% of the peak bandwidth under all traffic patterns. We conclude that given a fixed amount of input optical power, under realistic device assumptions, a point-to-point network has the best performance and energy characteristics.

Scaling trends of logic, memories, and interconnect networks lead towards dense many-core chips. Unfortunately, process yields and reticle sizes limit the scalability of large single-chip systems. Multi-chip systems break free of these areal limits, but in turn require enormous chip-to-chip bandwidth. The "macrochip" concept presented here integrates multiple many-core processor chips in a single package with silicon-photonic interconnects. This design enables a multi-chip system to approach the performance of a single large die.

We present a computing microsystem that uniquely leverages the bandwidth, density, and latency advantages of silicon photonic interconnect to enable highly compact supercomputer-scale systems. We describe and justify single-node and multinode systems interconnected with wavelength-routed optical links, quantify their benefits vis-a-vis electrically connected systems, analyze the constituent optical component and system requirements, and provide an overview of the critical technologies needed to fulfill this system vision. This vision calls for more than a hundredfold reduction in energy to communicate an optical bit of information. We explore the power dissipation of a photonic link, suggest a roadmap to lower the energy-per-bit of silicon photonic interconnects, and identify the challenges that will be faced by device and circuit designers towards this goal.

In this paper we explore the relationship between adder topology and energy efficiency. We compare the energy-delay tradeoff curves of selected 32- bit adder topologies, to determine how architectural features and design techniques affect energy efficiency. Optimizing different adders for the supply and threshold voltages, and transistor sizing, we show that topologies with the least number of logic stages having an average fanin of two per stage, and fewest wires are most energy efficient. While a design with fully custom sizes can be extremely tedious to layout, we show that custom sizing can be used as a guide to group different gates in the design, resulting in a manageable layout overhead without significant loss of energy efficiency.

Concern about the performance of wires wires in scaled technologies has led to research exploring other communication methods. This paper examines wire and gate delays as technologies migrate from 0.18-/spl mu/m to 0.035-/spl mu/m feature sizes to better understand the magnitude of the the wiring problem. Wires that shorten in length as technologies scale have delays that either track gate delays or grow slowly relative to gate delays. This result is good news since these "local" wires dominate chip wiring. Despite this scaling of local wire performance, computer-aided design (CAD) tools must still become move sophisticated in dealing with these wires. Under scaling, the total number of wires grows exponentially, so CAD tools will need to handle an ever-growing percentage of all the wires in order to keep designer workloads constant. Global wires present a more serious problem to designers. These are wires that do not scale in length since they communicate signals across the chip. The delay of these wives will remain constant if repeaters are used meaning that relative to gate delays, their delays scale upwards. These increased delays for global communication will drive architectures toward modular designs with explicit global latency mechanisms.

Trends in VLSI technology scaling demand that future computing devices be narrowly focused to achieve high performance and high efficiency, yet also target the high volumes and low costs of widely applicable general purpose designs. To address these conflicting requirements, we propose a modular reconfigurable architecture called Smart Memories, targeted at computing needs in the 0.1 /spl mu/m technology generation. A Smart Memories chip is made up of many processing tiles, each containing local memory, local interconnect, and a processor core. For efficient computation under a wide class of possible applications, the memories, the wires, and the computational model can all be altered to match the applications. To show the applicability of this design, two very different machines at opposite ends of the architectural spectrum, the Imagine stream processor and the Hydra speculative multiprocessor, are mapped onto the Smart Memories computing substrate. Simulations of the mappings show that the Smart Memories architecture can successfully map these architectures with only modest performance degradation.

Interconnect scaling to deep submicron processes presents many challenges to today's CAD flows. A recent analysis by D. Sylvester and K. Keutzer (1998) examined the behavior of average length wires under scaling, and controversially concluded that current CAD tools are adequate for future module-level designs. We show that average length wire scaling is sensitive to the technology assumptions, although the change in their behavior is small under all reasonable scaling assumptions. However, examining only average length wires is optimistic, since long wires are the ones that primarily cause CAD tool exceptions. In a module of fixed complexity, under both optimistic and pessimistic scaling assumptions, the number of long wires will increase slowly with scaling. More importantly, as the overall die capacity grows exponentially, the number of modules and thus the total number of wires in a design will also increase exponentially. Thus, if the design team size and per-designer workload is to remain relatively constant, future CAD tools will need to handle long wires much better than current tools to reduce the percentage of wires that require designer intervention.