We present a novel low latency CMOS hardware accelerator for fully connected (FC) layers in deep neural networks (DNNs). The FC accelerator, FC-ACCL, is based on 128 8x8 or 16x16 processing elements (PEs) for matrix-vector multiplication, and 128 multiply-accumulate (MAC) units integrated with 128 High Bandwidth Memory (HBM) units for storing the pretrained weights. Micro-architectural details for CMOS ASIC implementations are presented and simulated performance is compared to recent hardware accelerators for DNNs for AlexNet and VGG 16. When comparing simulated processing latency for a 4096-1000 FC8 layer, our FC-ACCL is able to achieve 48.4 GOPS (with a 100 MHz clock) which improves on a recent FC8 layer accelerator quoted at 28.8 GOPS with a 150 MHz clock. We have achieved this considerable improvement by fully utilizing the HBM units for storing and reading out column-specific FClayer weights in 1 cycle with a novel colum-row-column schedule, and implementing a maximally parallel datapath for processing these weights with the corresponding MAC and PE units. When up-scaled to 128 16x16 PEs, for 16x16 tiles of weights, the design can reduce latency for the large FC6 layer by 60 % in AlexNet and by 3 % in VGG16 when compared to an alternative EIE solution which uses compression.
University of Innsbruck（因斯布鲁克大学）；University of Erlangen-Nuremberg（埃尔朗根-纽伦堡大学）
A concept for a novel CMOS image sensor suited for analog image pre-processing is presented in this paper. As an example, an image restoration algorithm for reducing image noise is applied as image pre-processing in the analog domain. To supply low-latency data input for analog image preprocessing, the proposed concept for a CMOS image sensor offers a new sensor signal acquisition method in 2D. In comparison to image pre-processing in the digital domain, the proposed analog image pre-processing promises an improved image quality. Furthermore, the image noise at the stage of analog sensor signa…
To advance the energy efficiency of general digital computing far beyond the thermodynamic limits that apply to conventional digital circuits will require utilizing the principles of reversible computing. It has been known since the early 1990s that reversible computing based on adiabatic switching is possible in CMOS, although almost all of the "adiabatic" CMOS logic families in the literature are not actually fully adiabatic, which limits their achievable energy savings. The first CMOS logic style that achieved truly, fully adiabatic operation if leakage was negligible (CRL) is not fully sta…
For polar codes with short-to-medium code length, list successive cancellation decoding is used to achieve a good error-correcting performance. However, list pruning in the current list decoding is based on the sorting strategy and its timing complexity is high. This results in a long decoding latency for large list size. In this work, aiming at a low-latency list decoding implementation, a double thresholding algorithm is proposed for a fast list pruning. As a result, with a negligible performance degradation, the list pruning delay is greatly reduced. Based on the double thresholding, a low-…
Advanced Digital Sciences Centre（高级数字科学中心）；University of California, Santa Barbara（加州大学圣巴巴拉分校）；Georgia Institute of Technology（佐治亚理工学院）
The deep neural network (DNN) based AI applications on the edge require both low-cost computing platforms and high-quality services. However, the limited memory, computing resources, and power budget of the edge devices constrain the effectiveness of the DNN algorithms. Developing edge-oriented AI algorithms and implementations (e.g., accelerators) is challenging. In this paper, we summarize our recent efforts for efficient on-device AI development from three aspects, including both training and inference. First, we present on-device training with ultra-low memory usage. We propose a novel ran…
Herein, a bit-wise Convolutional Neural Network (CNN) in-memory accelerator is implemented using Spin-Orbit Torque Magnetic Random Access Memory (SOT-MRAM) computational sub-arrays. It utilizes a novel AND-Accumulation method capable of significantly-reduced energy consumption within convolutional layers and performs various low bit-width CNN inference operations entirely within MRAM. Power-intermittence resiliency is also enhanced by retaining the partial state information needed to maintain computational forward-progress, which is advantageous for battery-less IoT nodes. Simulation results i…
Due to their provably capacity-achieving performance, polar codes have attracted a lot of research interest recently. For a good error-correcting performance, list successive-cancellation decoding (LSCD) with large list size is used to decode polar codes. However, as the complexity and delay of the list management operation rapidly increase with the list size, the overall latency of LSCD becomes large and limits the applicability of polar codes in high-throughput and latency-sensitive applications. Therefore, in this work, the low-latency implementation for LSCD with large list size is studied…
Graphics Processing Units (GPUs) are currently the dominating programmable architecture for Deep Learning (DL) accelerators. The adoption of Field Programmable Gate Arrays (FPGAs) in DL accelerators is however getting momentum. In this paper, we demonstrate that Direct Hardware Mapping (DHM) of a Convolutional Neural Network (CNN) on an embedded FPGA substantially outperforms a GPU implementation in terms of energy efficiency and execution time. However, DHM is highly resource intensive and cannot fully substitute the GPU when implementing a state-of-the-art CNN. We thus propose a hybrid FPGA-…
Therapeutic intervention in neurological disorders still relies heavily on pharmacological solutions, while the treatment of patients with drug resistance remains an open challenge. This is particularly true for patients with epilepsy, 30% of whom are refractory to medications. Implantable devices for chronic recording and electrical modulation of brain activity have proved a viable alternative in such cases. To operate, the device should detect the relevant electrographic biomarkers from Local Field Potentials (LFPs) and determine the right time for stimulation. To enable timely interventions…
Limited by the memory capacity and compute power, singe-node graph convolutional neural network (GCN) accelerators cannot complete the execution of GCNs within a reasonable amount of time, due to the explosive size of graphs nowadays. Thus, large-scale GCNs call for a multi-node acceleration system (MultiAccSys) like TPU-Pod for large-scale neural networks. In this work, we aim to scale up single-node GCN accelerators to accelerate GCNs on large-scale graphs. We first identify the communication pattern and challenges of multi-node acceleration for GCNs on large-scale graphs. We observe that (1…
Neural networks are an increasingly attractive algorithm for natural language processing and pattern recognition. Deep networks with >50M parameters are made possible by modern GPU clusters operating at <50 pJ per op and more recently, production accelerators capable of <5pJ per operation at the board level. However, with the slowing of CMOS scaling, new paradigms will be required to achieve the next several orders of magnitude in performance per watt gains. Using an analog resistive memory (ReRAM) crossbar to perform key matrix operations in an accelerator is an attractive option. This work p…
Transformers have achieved great success in a wide variety of natural language processing (NLP) tasks due to the attention mechanism, which assigns an importance score for every word relative to other words in a sequence. However, these models are very large, often reaching hundreds of billions of parameters, and therefore require a large number of DRAM accesses. Hence, traditional deep neural network (DNN) accelerators such as GPUs and TPUs face limitations in processing Transformers efficiently. In-memory accelerators based on non-volatile memory promise to be an effective solution to this c…
This paper presents the Neural Cache architecture, which re-purposes cache structures to transform them into massively parallel compute units capable of running inferences for Deep Neural Networks. Techniques to do in-situ arithmetic in SRAM arrays, create efficient data mapping and reducing data movement are proposed. The Neural Cache architecture is capable of fully executing convolutional, fully connected, and pooling layers in-cache. The proposed architecture also supports quantization in-cache. Our experimental results show that the proposed architecture can improve inference latency by 1…
A Low-Latency ASIC Architecture for Real-Time Line Segment Detection
School of Computer Engineering, Iran University of Science and Technology（伊朗理工大学计算机工程学院）；Electrical, Computer, and Systems Engineering Department, Case Western Reserve University（凯斯西储大学电气、计算机和系统工程系）
Computationally intensive Inference tasks of Deep neural networks have enforced revolution of new accelerator architecture to reduce power consumption as well as latency. The key figure of merit in hardware inference accelerators is the number of multiply-and-accumulation operations per watt (MACs/W), where, the state-of-the-arts MACs/W remains several hundreds Giga-MACs/W. We propose a Tera-MACS/W neural hardware inference Accelerator (TMA) with 8-bit activations and scalable integer weights less than 1-byte. The architectures main feature is configurable neural processing element for matrix-…
End-to-End Transformer Acceleration Through Processing-in-Memory Architectures
Department of Electrical and Computer Engineering, University of Virginia（弗吉尼亚大学电气与计算机工程系）；Department of Electrical and Computer Engineering, Duke University（杜克大学电气与计算机工程系）
In recent years, neural networks have surpassed classical algorithms in areas such as object recognition, e.g. in the well-known ImageNet challenge. As a result, great effort is being put into developing fast and efficient accelerators, especially for Convolutional Neural Networks (CNNs). In this work we present ConvAix, a fully C-programmable processor, which -- contrary to many existing architectures -- does not rely on a hard-wired array of multiply-and-accumulate (MAC) units. Instead it maps computations onto independent vector lanes making use of a carefully designed vector instruction se…
University of Nevada, Las Vegas（内华达大学拉斯维加斯分校）；Southern China Institute of Technology（华南理工学院）
The increasing popularity of deep neural network (DNN) applications demands high computing power and efficient hardware accelerator architecture. DNN accelerators use a large number of processing elements (PEs) and on-chip memory for storing weights and other parameters. As the communication backbone of a DNN accelerator, networks-on-chip (NoC) play an important role in supporting various dataflow patterns and enabling processing with communication parallelism in a DNN accelerator. However, the widely used mesh-based NoC architectures inherently cannot support the efficient one-to-many and man…
Previous graph analytics accelerators have achieved great improvement on throughput by alleviating irregular off-chip memory accesses. However, on-chip side datapath conflicts and design centralization have become the critical issues hindering further throughput improvement. In this paper, a general solution, Multiple-stage Decentralized Propagation network (MDP-network), is proposed to address these issues, inspired by the key idea of trading latency for throughput. Besides, a novel High throughput Graph analytics accelerator, HiGraph, is proposed by deploying MDP-network to address each issu…
Deep learning has led to unprecedented successes in solving some very difficult problems in domains such as computer vision, natural language processing, and general pattern recognition. These achievements are the culmination of decades-long research into better training techniques and deeper neural network models, as well as improvements in hardware platforms that are used to train and execute the deep neural network models. Many application-specific integrated circuit (ASIC) hardware accelerators for deep learning have garnered interest in recent years due to their improved performance and e…
Recent years have seen a rapid increase in research activity in the field of DRAM-based Processing-In-Memory (PIM) accelerators, where the analog computing capability of DRAM is employed by minimally changing the inherent structure of DRAM peripherals to accelerate various data-centric applications. Several DRAM-based PIM accelerators for Convolutional Neural Networks (CNNs) have also been reported. Among these, the accelerators leveraging in-DRAM stochastic arithmetic have shown manifold improvements in processing latency and throughput, due to the ability of stochastic arithmetic to convert …
Network-on-Chip (NoC) plays a significant role in the performance of a DNN accelerator. The scalability and modular design property of the NoC help in improving the performance of a DNN execution by providing flexibility in running different kinds of workloads. Data movement in a DNN workload is still a challenging task for DNN accelerators and hence a novel approach is required. In this paper, we propose the In-Network Accumulation (INA) method to further accelerate a DNN workload execution on a many-core spatial DNN accelerator for the Weight Stationary (WS) dataflow model. The INA method ex…
Transformer-based language models such as BERT provide significant accuracy improvement for a multitude of natural language processing (NLP) tasks. However, their hefty computational and memory demands make them challenging to deploy to resource-constrained edge platforms with strict latency requirements. We present EdgeBERT, an in-depth algorithm-hardware co-design for latency-aware energy optimization for multi-task NLP. EdgeBERT employs entropy-based early exit predication in order to perform dynamic voltage-frequency scaling (DVFS), at a sentence granularity, for minimal energy consumption…
FPGA-based Acceleration for Convolutional Neural Networks: A Comprehensive Review
Key Laboratory of Low-Dimensional Quantum Structures and Quantum Control of Ministry of Education, School of Physics and Electronics, Hunan Normal University（教育部低维量子结构与量子控制重点实验室，物理电子学院，湖南师范大学）
Intensive computation is entering data centers with multiple workloads of deep learning. To balance the compute efficiency, performance, and total cost of ownership (TCO), the use of a field-programmable gate array (FPGA) with reconfigurable logic provides an acceptable acceleration capacity and is compatible with diverse computation-sensitive tasks in the cloud. In this paper, we develop an FPGA acceleration platform that leverages a unified framework architecture for general-purpose convolutional neural network (CNN) inference acceleration at a data center. To overcome the computation bound,…
