The adoption of autonomous swarms, consisting of a multitude of unmanned aerial vehicles (UAVs), operating in a collaborative manner, has become prevalent in mainstream application domains for both military and civilian purposes. These swarms are expected to collaboratively carry out navigation tasks and employ complex reinforcement learning (RL) models within the stringent onboard size, weight, and power constraints. While techniques such as reducing onboard operating voltage can improve the energy efficiency of both computation and flight missions, they can lead to on-chip bit failures that are detrimental to mission safety and performance.

Deep neural networks (DNNs) have become ubiquitous in machine learning, but their energy consumption remains problematically high. An effective strategy for reducing such consumption is supply-voltage reduction, but if done too aggressively, it can lead to accuracy degradation. This is due to random bit-flips in static random access memory (SRAM), where model parameters are stored. To address this challenge, we have developed NeuralFuse, a novel add-on module that handles the energy-accuracy tradeoff in low-voltage regimes by learning input transformations and using them to generate error-resistant data representations, thereby protecting DNN accuracy in both nominal and low-voltage scenarios. As well as being easy to implement, NeuralFuse can be readily applied to DNNs with limited access, such cloud-based APIs that are accessed remotely or non-configurable hardware. Our experimental results demonstrate that, at a 1% bit-error rate, NeuralFuse can reduce SRAM access energy by up to 24% while recovering accuracy by up to 57%. To the best of our knowledge, this is the first approach to addressing low-voltage-induced bit errors that requires no model retraining.

Autonomous systems, such as Unmanned Aerial Vehicles (UAVs), are expected to run complex reinforcement learning (RL) models to execute fully autonomous position-navigation-time tasks within stringent onboard weight and power constraints. We observe that reducing onboard operating voltage can benefit the energy efficiency of both the computation and flight mission, however, it can also result in on-chip bit failures that are detrimental to mission safety and performance. To this end, we propose BERRY, a robust learning framework to improve bit error robustness and energy efficiency for RL-enabled autonomous systems. BERRY supports robust learning, both offline and on-board the UAV, and for the first time, demonstrates the practicality of robust low-voltage operation on UAVs that leads to high energy savings in both compute-level operation and system-level quality-of-flight. We perform extensive experiments on 72 autonomous navigation scenarios and demonstrate that BERRY generalizes well across environments, UAVs, autonomy policies, operating voltages and fault patterns, and consistently improves robustness, efficiency and mission performance, achieving up to 15.62% reduction in flight energy, 18.51% increase in the number of successful missions, and 3.43× processing energy reduction.

Deep neural network (DNN) accelerators received considerable attention in recent years due to the potential to save energy compared to mainstream hardware. Low-voltage operation of DNN accelerators allows to further reduce energy consumption, however, causes bit-level failures in the memory storing the quantized weights. Furthermore, DNN accelerators are vulnerable to adversarial attacks on voltage controllers or individual bits. In this paper, we show that a combination of robust fixed-point quantization, weight clipping, as well as random bit error training (RandBET) or adversarial bit error training (AdvBET) improves robustness against random or adversarial bit errors in quantized DNN weights significantly. This leads not only to high energy savings for low-voltage operation as well as low-precision quantization, but also improves security of DNN accelerators. In contrast to related work, our approach generalizes across operating voltages and accelerators and does not require hardware changes. Moreover, we present a novel adversarial bit error attack and are able to obtain robustness against both targeted and untargeted bit-level attacks. Without losing more than 0.8%/2% in test accuracy, we can reduce energy consumption on CIFAR10by 20%/30% for 8/4-bit quantization. Allowing up to 320 adversarial bit errors, we reduce test error from above 90% (chance level) to 26.22%.

We present a scalable methodology for the agile physical design of tile-based heterogeneous system-on-chip (SoC) architectures that simplifies the reuse and integration of open-source hardware components. The methodology leverages the regularity of the on-chip communication infrastructure, which is based on a multi-plane network-on-chip (NoC), and the modularity of socket interfaces, which connect the tiles to the NoC. Each socket also provides its tile with a set of platform services, including independent clocking and voltage control. As a result, the physical design of each tile can be decoupled from its location in the top-level floorplan of the SoC and the overall SoC design can benefit from a hierarchical timing-closure flow, design reuse and, if necessary, fast respin. With the proposed methodology we completed two SoC tapeouts of increasing complexity, which illustrate its capabilities and the resulting gains in terms of design productivity.

Autonomous systems have reached a tipping point, with a myriad of self-driving cars, unmanned aerial vehicles (UAVs), and robots being widely applied and revolutionizing new applications. The continuous deployment of autonomous systems reveals the need for designs that facilitate increased resiliency and safety. The ability of an autonomous system to tolerate, or mitigate against errors, such as environmental conditions, sensor, hardware and software faults, and adversarial attacks, is essential to ensure its functional safety. Application-aware resilience metrics, holistic fault analysis frameworks, and lightweight fault mitigation techniques are being proposed for accurate and effective resilience and robustness assessment and improvement. This paper explores the origination of fault sources across the computing stack of autonomous systems, discusses the various fault impacts and fault mitigation techniques of different scales of autonomous systems, and concludes with challenges and opportunities for assessing and building next-generation resilient and robust autonomous systems.

Federated learning (FL) is a machine learning paradigm that enables a cluster of decentralized edge devices to collaboratively train a shared machine learning model without exposing users' raw data. However, the intensive model training computation is energy-demanding and poses severe challenges to end devices' battery life. In this paper, we present BoFL, a training pace controller deployed on the edge devices that actuates the hardware operational frequencies over multiple configurations to achieve energy-efficient federated learning. BoFL operates in an explore-then-exploit manner within limited rounds of FL tasks. BoFL explores the large hardware frequency space strategically with a tailor-designed Bayesian optimization algorithm. BoFL first finds a set of good operational configurations within few task training rounds, and then exploits these configurations in the remaining rounds to achieve minimized energy consumption for model training. Experiments on multiple real-world edge devices with different FL tasks suggest that BoFL can reduce energy consumption of model training by around 26%, and achieve near-optimal energy efficiency.

Graph data structures are central to many applications such as social networks, citation networks, molecular interactions, and navigation systems. Graph Convolutional Networks (GCNs) are used to process and learn insights from the graph data for tasks such as link prediction, node classification, and learning node embeddings. The compute and memory access characteristics of GCNs differ, both from conventional graph analytics algorithms and from convolutional neural networks, rendering the existing accelerators for graph analytics as well as deep learning, inefficient. In this work, we propose PIM-GCN, a crossbar-based processing-in-memory (PIM) accelerator architecture for GCNs. PIM-GCN incorporates a node-stationary dataflow with support for both Compressed Sparse Row (CSR) and Compressed Sparse Column (CSC) graph data representations. We propose techniques for graph traversal in the compressed sparse domain, feature aggregation, and feature transformation operations in GCNs mapped to in-situ analog compute functions of crossbar memory, and present the trade-offs in performance, energy, and scalability aspects of the PIM-GCN architecture for CSR, and CSC graph data representations. PIM-GCN shows an average speedup of over$3-16\times$and an average energy reduction of$4-12\times$compared to the existing accelerator architectures.

We present the novel micro-architectural features, supported by an innovative and novel pre-silicon methodology in the design of POWER10. The resulting projected energy efficiency boost over POWER9 is 2.6x at core level (for SPECint) and up to 3x at socket level. In addition, a new feature supporting inline AI acceleration was added to the POWER ISA and incorporated into the POWER10 processor core design. The resulting boost in SIMD/AI socket performance is projected to be up to 10x for FP32 and 21x for INT8 models of ResNet-50 and BERT-Large. In this paper, we describe the novel methodology deployed and used not only to obtain these efficiency boosts for traditional workloads, but also to infuse AI/ML/HPC capability directly into the POWER10 core.

Graph analytics applications are ubiquitous in this era of a connected world. These applications have very low compute to byte-transferred ratios and exhibit poor locality, which limits their computational efficiency on general purpose computing systems. Conventional hardware accelerators employ custom dataflow and memory hierarchy organization to overcome these challenges. Processing-in-memory (PIM) accelerators leverage massively parallel compute capable memory arrays to perform the in-situ operations on graph data or employ custom compute elements near the memory to leverage larger internal bandwidths. In this work, we present GaaS-X, a graph analytics accelerator that inherently supports the sparse graph data representations using an in-situ compute-enabled crossbar memory architectures. We alleviate the overheads of redundant writes, sparse to dense conversions, and redundant computations on the invalid edges that are present in the state of the art crossbar-based PIM accelerators. GaaS-X achieves 7.7× and 2.4× performance and 22× and 5.7×, energy savings, respectively, over two state-of-the-art crossbar accelerators and offers orders of magnitude improvements over GPU and CPU solutions.

Low voltage architecture and design are key enablers of high throughput per watt in heterogeneous, accelerator-rich many-core designs. However, such low voltage operation poses significant challenges because of difficulties in achieving reliable functionality of on-chip memories, particularly SRAMs at these design points. In this paper, we present a technique of low-voltage neural network acceleration, where the embedded SRAM architecture is equipped with a novel application-aware supply voltage boosting capability. This technique mitigates low-voltage induced failures, while enabling Very low voltage (VLV)1operation during most of the application run, resulting in substantial improvement in net energy efficiency. We present a framework to evaluate the impact of low-voltage SRAM errors on machine learning applications and characterize trade-offs between output inference accuracy and energy efficiency in our application-programmable supply boosted SRAM architecture. Using the proposed technique we push the limits on the minimum operable voltage (Vmin) for the desired output quality. As a proof of concept, we demonstrate these techniques on Dante, a Deep Neural Network (DNN) accelerator chip taped out in state-of-the art 14nm technology.

Defining a processor micro-architecture for a targeted productspace involves multi-dimensional optimization across performance, power and reliability axes. A key decision in sucha definition process is the circuit-and technology-driven parameterof the nominal (voltage, frequency) operating point. This is a challenging task, since optimizing individually orpair-wise amongst these metrics usually results in a designthat falls short of the specification in at least one of the threedimensions. Aided by academic research, industry has nowadopted early-stage definition methodologies that considerboth energy-and performance-related metrics. Reliabilityrelatedenhancements, on the other hand, tend to get factoredin via a separate thread of activity. This task is typically pursuedwithout thorough pre-silicon quantifications of the energyor even the performance cost. In the late-CMOS designera, reliability needs to move from a post-silicon afterthoughtor validation-only effort to a pre-silicon definitionprocess. In this paper, we present BRAVO, a methodologyfor such reliability-aware design space exploration. BRAVOis supported by a multi-core simulation framework that integratesperformance, power and reliability modeling capability. Errors induced by both soft and hard fault incidence arecaptured within the reliability models. We introduce the notionof the Balanced Reliability Metric (BRM), that we useto evaluate overall reliability of the processor across soft andhard error incidences. We demonstrate up to 79% improvementin reliability in terms of this metric, for only a 6% dropin overall energy efficiency over design points that maximizeenergy efficiency. We also demonstrate several real-life usecaseapplications of BRAVO in an industrial setting.

This paper is a tutorial-style introduction to a special session on: Effective Voltage Scaling in the Late CMOS Era. It covers the fundamental challenges and associated solution strategies in pursuing very low voltage (VLV) designs. We discuss the performance and system reliability constraints that are key impediments to VLV. The associated trade-offs across power, performance and reliability are helpful in inferring the optimal operational voltage-frequency point. This work was performed under the auspices of an ongoing DARPA program (named PERFECT) that is focused on maximizing system-level energy efficiency.

Limited power budgets and the need for high performance computing have led to platform customization with a number of accelerators integrated with CMPs. In order to study customized architectures, we model four customization design points and compare their performance and energy across a number of computer vision workloads. We analyze the limitations of generic architectures and quantify the costs of increasing customization using these micro-architectural design points. This analysis leads us to develop a framework consisting of low-power multi-cores and an array of configurable micro-accelerator functional units. Using this platform, we illustrate dataflow and control processing optimizations that provide for performance gains similar to custom ASICs for a wide range of vision benchmarks.

Steep Slope devices, with Heterojunction Tunnel FETs (TFETs) in particular, have been proposed as a viable solution to overcome the subthreshold slope limitation in existing CMOS technology and achieve ultra-low voltage operation with acceptable performance. However, state-of-the-art FinFET technologies continue to demonstrate superior performance than steep slope devices in application domains demanding peak single threaded performance. In this context, we examine different computing paradigms where TFET technologies can be used, not just as a `drop in' replacement, but as an additional parameter to augment the architectural design space. This greatly widens the scope of optimizations for performance and power. We investigate the tradeoffs between device and architectures in general purpose processors when performance, power and temperature are individually constrained. We also synthesize examples of domain-specific accelerators used in computer vision using in-house TFET standard cell libraries to demonstrate the energy benefits of designing TFET-based accelerators. We demonstrate that synthesizing these accelerators using TFETs reduces energy by over 6X in comparison to an equivalent iso-voltage CMOS-based design and by over 30% in comparison to an iso-performance CMOS design.

Video applications are becoming ubiquitous in mobile and embedded systems. Wearable video systems such as Google Glasses require capabilities for real-time video analytics and prolonged battery lifetimes. Further, the increasing resolution of image sensors in these mobile systems places an increasing demand on both the memory storage as well as the computational power. In this work, we present the Refresh Enabled Video Analytics (REVA) system, an embedded architecture for multi-object scene understanding and tackle the unique opportunities provided by real-time embedded video analytics applications for reducing the DRAM memory refresh energy. We compare our design with the existing design space and show savings of 88% in refresh power and 15% in total power, as compared to a standard DRAM refresh scheme.

Summary form only given. The objective of this paper is to present a configurable architecture for a visual saliency model based on AIM. It presents algorithmic enhancements to AIM that facilitates the design of a performance-efficient hardware architecture that offers tradeoffs between accuracy, resource utilization and latency. The AIM computational model involves (1) extraction of a set of coefficient features for each local patch in an image, (2) estimation of probability density for each coefficient with respect to its local surround, (3) computation of their product to give a joint likelihood and (4) computation of the self information of each pixel from its log likelihood. Calculation of likelihood with respect to each pixel individually in a local surround is computationally expensive. It proposes to approximate the contribution of pixels in the surround in terms of “cells” grouped further into “support zones”, whose widths are configurable. This approximation leads to nearly a 10x reduction in the number of multipliers, a critical resource, for a 41x41 surround size.

Video analytics introduce new levels of intelligence to automated scene understanding. Neuromorphic algorithms, such as HMAX, are proposed as robust and accurate algorithms that mimic the processing in the visual cortex of the brain. HMAX, for instance, is a versatile algorithm that can be repurposed to target several visual recognition applications. This paper presents the design and evaluation of hardware accelerators for extracting visual features for universal recognition. The recognition applications include object recognition, face identification, facial expression recognition, and action recognition. These accelerators were validated on a multi-FPGA platform and significant performance enhancement and power efficiencies were demonstrated when compared to CMP and GPU platforms. Results demonstrate as much as 7.6X speedup and 12.8X more power-efficient performance when compared to those platforms.