Eyeriss: An Energy-Efficient Reconfigurable Accelerator for Deep Convolutional Neural Networks
The DOI matches its Crossref record; 38 of the 42 references it cites were found in a registry or an index.
Abstract
Eyeriss is an accelerator for state-of-the-art deep convolutional neural networks (CNNs). It optimizes for the energy efficiency of the entire system, including the accelerator chip and off-chip DRAM, for various CNN shapes by reconfiguring the architecture. CNNs are widely used in modern AI systems but also bring challenges on throughput and energy efficiency to the underlying hardware. This is because its computation requires a large amount of data, creating significant data movement from on-chip and off-chip that is more energy-consuming than computation. Minimizing data movement energy cost for any CNN shape, therefore, is the key to high throughput and energy efficiency. Eyeriss achieves these goals by using a proposed processing dataflow, called row stationary (RS), on a spatial architecture with 168 processing elements. RS dataflow reconfigures the computation mapping of a given shape, which optimizes energy efficiency by maximally reusing data locally to reduce expensive data movement, such as DRAM accesses. Compression and data gating are also applied to further improve energy efficiency. Eyeriss processes the convolutional layers at 35 frames/s and 0.0029 DRAM access/multiply and accumulation (MAC) for AlexNet at 278 mW (batch size N = 4), and 0.7 frames/s and 0.0035 DRAM access/MAC for VGG-16 at 236 mW (N = 3).
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“such scaling will enable NeuRRAM to deliver higher energy and area efficiency than today’s state-of-the-art edge inference accelerators58–61.”
From A compute-in-memory chip based on resistive random-access memory (2022) · cites this paper as [58]
In the citing paper
“Some studies proposed to implement the argmax() function via hardware128–148, which could be beneficial to reduce the total transistor count and power consumption while at the same time increasing the throughput.”
From Hardware implementation of memristor-based artificial neural networks (2024) · cites this paper as [131]
In the citing paper
“Based on the analysis of the research, CoNNA is able to speed up at least 3.08 times and up to 14.1 times faster than the MIT Eyeriss Accelerator [20].”
From Embedded Intelligence on FPGA: Survey, Applications and Challenges (2021) · cites this paper as [20]
In the citing paper
“massively parallel processing architectures with distributed memory architectures, such as graphic processing units (GPUs) but also increasingly specialized chips for NN inference (e.g., [17]), have shown to be very energy efficient and capable of impressive computational throughput when fully utilized by concurrent algorithms [18].”
From An Introduction to Deep Learning for the Physical Layer (2017) · cites this paper as [17]
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 1×
“Modern CNNs achieve their superior accuracy by building a very deep hierarchy of layers [2]–[5], [7], which transform the input image data into highly abstract representations called feature maps (fmaps).”
BackgroundVery Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148 · cited 7×
“We benchmark the chip performance using two publicly available and widely used CNNs: AlexNet [2] and VGG-16 [3].”
MethodInfluentialImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155
LeCun, Y., Bengio, Y., & Hinton, G. (2015) · Nature · doi:10.1038/nature14539 · Cited by 84,126
Gradient-based learning applied to document recognition
Lecun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998) · Proceedings of the IEEE · doi:10.1109/5.726791 · Cited by 58,805 · cited 1×
“[25] demonstrate a CNN processor and only report the theoretical peak throughput along with the power measured on a CNN for the MNIST data set [26], which has storage and computation requirements that are orders of magnitude lower than the state-of-the-art CNNs.”
MethodGoing deeper with convolutions
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298594 · Cited by 48,562
ImageNet Large Scale Visual Recognition Challenge
Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. (2015) · International Journal of Computer Vision · doi:10.1007/s11263-015-0816-y · Cited by 39,380 · cited 2×
“We have demonstrated an 1000-class image classification task [27] using this system, and a live demo can be found in [29].”
BackgroundMethodRich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation
Girshick, R., Donahue, J., Darrell, T., & Malik, J. (2014) · 2014 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2014.81 · Cited by 32,684
Mastering the game of Go with deep neural networks and tree search
Silver, D., Huang, A., Maddison, C. J., Guez, A., Sifre, L., van den Driessche, G., Schrittwieser, J., Antonoglou, I., Panneershelvam, V., Lanctot, M., et al. (2016) · Nature · doi:10.1038/nature16961 · Cited by 15,612 · cited 1×
“D EEP learning using convolutional neural networks (CNNs) [1] has achieved unprecedented accuracy on many modern AI applications [2]–[9].”
Background(2010) · International Conference on Machine Learning · Cited by 13,995
Research works printed but not linked: 2
- han, 2015, Learning both weights and connections for efficient neural network, Proc Adv Neural Inf Process Syst, 28, 1135
- chen, 2016, An 1000-Class Image Classification Task Performed by the Eyeriss-Integrated Deep Learning System
Cited by
Items 901–1,000 of 1,001
EdgeL^3: Compressing L^3-Net for Mote Scale Urban Noise Monitoring
Kumari, S., Roy, D., Cartwright, M., Bello, J. P., & Arora, A. (2019) · IEEE International Symposium on Parallel & Distributed Processing, Workshops and Phd Forum · doi:10.1109/ipdpsw.2019.00145 · Cited by 18
Escoin: Efficient Sparse Convolutional Neural Network Inference on GPUsPreprint
Chen, X. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1802.10280 · Cited by 33
Marvel: A Data-centric Compiler for DNN Operators on Spatial AcceleratorsPreprint
Chatarasi, P., Kwon, H., Raina, N., Malik, S., Haridas, V., Parashar, A., Pellauer, M., Krishna, T., & Sarkar, V. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2002.07752 · Cited by 28
“…and energy, accounting for under-utilization, edge conditions, and data reuse or movement across time (via L1/L2 buffers [7]), space (via broadcast links [24]), and space-time (via neighboring links [8, 18]), and (iii) light-weight and fast to enable rapid evaluation through a large search space.”
BackgroundSelf-citationCoNNa–Hardware accelerator for compressed convolutional neural networks
Struharik, R. J. R., Vukobratović, B. Z., Erdeljan, A. M., & Rakanović, D. M. (2020) · Microprocessors and Microsystems · doi:10.1016/j.micpro.2020.102991 · Cited by 18
Neuromorphic Computing with Memcapacitors: Advancements, Challenges, and Future Directions
AbuHamra, N., Khan, M. U., Hassan, E., Al‐Qutayri, M., & Mohammad, B. S. (2025) · Advanced Electronic Materials · doi:10.1002/aelm.202500250 · Cited by 21
Lee, J., Shin, D., & Yoo, H. (2017) · IEEE Asian Solid-State Circuits Conference (A-SSCC) · doi:10.1109/asscc.2017.8240260 · Cited by 18
Introduction to hardware accelerator systems for artificial intelligence and machine learning
Gupta, N. (2020) · Advances in computers · doi:10.1016/bs.adcom.2020.07.001 · Cited by 20
Transmuter: Bridging the Efficiency Gap using Memory and Dataflow Reconfiguration
Pal, S., Feng, S., Park, D.-H., Kim, S., Amarnath, A., Yang, C.-S., He, X. J., Beaumont, J., May, K., Xiong, Y., et al. (2020) · International Conference on Parallel Architectures and Compilation Techniques (PACT) · doi:10.1145/3410463.3414627 · Cited by 30
High-throughput Near-Memory Processing on CNNs with 3D HBM-like Memory
Park, N., Ryu, S., Kung, J., & Kim, J. (2021) · ACM Transactions on Design Automation of Electronic Systems · doi:10.1145/3460971 · Cited by 25
Modeling and Optimization of SRAM-based In-Memory Computing Hardware Design
Saikia, J., Yin, S., Cherupally, S. K., Zhang, B., Meng, J., Seok, M., & Seo, J.-S. (2021) · Design, Automation, and Test in Europe · doi:10.23919/date51398.2021.9473973 · Cited by 21
Chen, X., Yajima, T., Inoue, I. H., & Iizuka, T. (2022) · Japanese Journal of Applied Physics · doi:10.35848/1347-4065/ac43e4 · Cited by 20
Client-optimized algorithms and acceleration for encrypted compute offloading
van der Hagen, M., & Lucia, B. (2022) · International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS) · doi:10.1145/3503222.3507737 · Cited by 18
“Client DNN inference performance is improving through software [1, 52] and hardware optimization [12, 30].”
MethodNekooei, A., & Safari, S. (2022) · Neural Networks · doi:10.1016/j.neunet.2022.02.024 · Cited by 22
Memristor-based Deep Spiking Neural Network with a Computing-In-Memory Architecture
Nowshin, F., & Yi, Y. (2022) · International Symposium on Quality Electronic Design (ISQED) · doi:10.1109/isqed54688.2022.9806206 · Cited by 18
Efficiency analysis of artificial vs. Spiking Neural Networks on FPGAs
Li, Z., Lemaire, E., Abderrahmane, N., Bilavarn, S., & Miramond, B. (2022) · Journal of Systems Architecture · doi:10.1016/j.sysarc.2022.102765 · Cited by 23
Liu, Q., Setter, J., Huff, D., Strange, M., Feng, K., Horowitz, M., Raina, P., & Kjølstad, F. (2022) · ACM Transactions on Architecture and Code Optimization · doi:10.1145/3572908 · Cited by 22
DeepFlow: A Cross-Stack Pathfinding Framework for Distributed AI Systems
Ardalani, N., Pal, S., & Gupta, P. (2023) · ACM Transactions on Design Automation of Electronic Systems · doi:10.1145/3635867 · Cited by 27
“The effective bandwidth for each link is downrated by the number of logical edges sharing the link.”
BackgroundAutomated Deep Learning: Neural Architecture Search Is Not the End
Dong, X., Kedziora, D. J., Musiał, K., & Gabryś, B. (2024) · Foundations and Trends® in Machine Learning · doi:10.1561/2200000119 · Cited by 53
Das, A., Russo, E., & Palesi, M. (2024) · IEEE Transactions on Computers · doi:10.1109/tc.2024.3386067 · Cited by 28
Suleiman, A., Zhang, Z., & Sze, V. (2017) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2017.2648820 · Cited by 20
Self-citationPower Reduction in CNN Pooling Layers with a Preliminary Partial Computation Strategy
Ahmadi, M., Vakili, S., Langlois, J. M. P., & Gross, W. J. (2018) · IEEE International New Circuits and Systems Conference (NEWCAS) · doi:10.1109/newcas.2018.8585433 · Cited by 22
“[7] proposed a high accuracy accelerator for convolutional layers of AlexNet where lowpower techniques such as zero skipping and clock gating are used to prevent unnecessary computations.”
MethodSinangil, M. E., Lin, Y., Liao, H.-J., & Chang, J. T.-Y. (2019) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2019.2895236 · Cited by 18
“For example, Eyeriss [3] accelerator chip employs a 108-kB global buffer, while each one of the 168 processing elements contains a 3.”
MethodInfluentialHyperdrive: A Multi-Chip Systolically Scalable Binary-Weight CNN Inference Engine
Andri, R., Cavigelli, L., Rossi, D., & Benini, L. (2019) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2019.2905654 · Cited by 19
Yoon, I., Anwar, M. A., Joshi, R., Rakshit, T., & Raychowdhury, A. (2019) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2019.2932285 · Cited by 23
Enabling Timing Error Resilience for Low-Power Systolic-Array Based Deep Learning Accelerators
Zhang, J. J., Ghodsi, Z., Garg, S., & Rangineni, K. (2019) · IEEE Design and Test · doi:10.1109/mdat.2019.2947271 · Cited by 18
“Another common technique to enhance DNN efficiency is run-time pruning [1]; in the “Experimental results” section, we show that ThunderVolt is synergistic with Zero-Skipping, a commonly used instantiation of run-time pruning.”
MethodInfluentialSwallow: A Versatile Accelerator for Sparse Neural Networks
Liu, B., Chen, X., Han, Y., & Xu, H. (2020) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2020.2978836 · Cited by 19
“For example, Diannao [15] and Eyeriss [16], [21] focus on general neural networks without supporting the sparsity in computation.”
BackgroundLeng, J., Buyuktosunoglu, A., Bertran, R., Bose, P. K., Chen, Q., Guo, M., & Reddi, V. J. (2020) · IEEE International Symposium on High-Performance Computer Architecture (HPCA) · doi:10.1109/hpca47549.2020.00014 · Cited by 18
“, a convolutional layer accelerator [36] takes two input memory regions (for weight and input feature map) and stores the output feature map in another memory region.”
BackgroundVerhelst, M., & Murmann, B. (2020) · The frontiers collection · doi:10.1007/978-3-030-18338-7_18 · Cited by 23
RAMANN: in-SRAM differentiable memory computations for memory-augmented neural networks
Ali, M., Agrawal, A., & Roy, K. (2020) · International Symposium on Low Power Electronics and Design (ISLPED) · doi:10.1145/3370748.3406574 · Cited by 20
“Standard DNN accelerators are originally built for either convolutional neural network (CNNs) operations that are compute-bound, or fully-connected networks that are memory-bound and can be accelerated by input-batching and weight-reuse [ 11, 12].”
FPGA based neural network accelerators
Kim, J.-Y. (2020) · Advances in computers · doi:10.1016/bs.adcom.2020.11.002 · Cited by 22
Edge Artificial Intelligence Chips for the Cyberphysical Systems Era
Fuketa, H., & Uchiyama, K. (2021) · Computer · doi:10.1109/mc.2020.3034951 · Cited by 18
Artificial Intelligence Accelerators Based on Graphene Optoelectronic Devices
Gao, W., Yu, C., & Chen, R. (2021) · Advanced Photonics Research · doi:10.1002/adpr.202100048 · Cited by 23
Specializing CGRAs for Light-Weight Convolutional Neural Networks
Lee, J., & Lee, J. (2021) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2021.3123178 · Cited by 18
“To analyze performance bottleneck of CGRAs, we compare a baseline CGRA [6] with Eyeriss [16], a reference hard DPU, in terms of minimum theoretical latency, using seven DWC layers from MobileNet V2, one from each bottleneck (we see similar results with other layers as well).”
BackgroundMethodInfluentialSakthivel, R., Thaseen, I. S., Vanitha, M., Deepa, M., Angulakshmi, M., Ramaiah, M., Mahendran, A., Alnumay, W. S., & Chatterjee, P. (2022) · Sustainable Cities and Society · doi:10.1016/j.scs.2022.103713 · Cited by 21
“Most of the companies like Intel, Google, and Facebook, etc. started exploring and using the AI algorithms and hardware architectures to the great extent (Chen, Krishna, Emer, & Sze, 2016).”
A hardware-efficient computing engine for FPGA-based deep convolutional neural network accelerator
Li, X., Huang, H., Chen, T., Gao, H., Hu, X., & Xiong, X. (2022) · Microelectronics Journal · doi:10.1016/j.mejo.2022.105547 · Cited by 18
Edge ai: leveraging the full potential of deep learning
Shuvo, M. M. H. (2022) · Studies in Computational Intelligence · doi:10.1007/978-3-031-14748-7_2 · Cited by 18
FPAP: A Folded Architecture for Energy-Quality Scalable Convolutional Neural Networks
Wang, Y., Lin, J. T., & Wang, Z. (2018) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2018.2856624 · Cited by 22
“Many prior CNN architectures use bit-parallel 16 bits MAC units [4], [16], [28], which is inefficient when encountering mixed-precision models.”
MethodAccelerating Low Bit-Width Deep Convolution Neural Network in MRAM
He, Z., Angizi, S., & Fan, D. (2018) · IEEE Computer Society Annual Symposium on VLSI · doi:10.1109/isvlsi.2018.00103 · Cited by 20
“In comparison with normal network accelerator design [25] using 32bit floating multiplication (i.”
Compact Convolutional Neural Network Accelerator for IoT Endpoint SoC
Ge, F., Wu, N., Xiao, H., Zhang, Y., & Zhou, F. (2019) · Electronics · doi:10.3390/electronics8050497 · Cited by 21
Zhu, M., Liu, L., Wang, C., & Xie, Y. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1606.06234 · Cited by 27
Delmás, A., Judd, P., Stuart, D. M., Poulos, Z., Mahmoud, M., Sharify, S., Nikolić, M., & Moshovos, A. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1803.03688 · Cited by 25
“In practice some weights are zero and some activations are zero (or close enough to zero that can be treated as such), and hence are ineffectual [3, 4, 7, 20, 24].”
BackgroundInfluentialHanif, M. A., Putra, R. V. W., Tanvir, M., Hafiz, R., Rehman, S., & Shafique, M. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1810.12910 · Cited by 24
An Introduction to Deep Learning for the Physical Layer
O’Shea, T. J., & Hoydis, J. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1702.00832 · Cited by 15
“, [17]), have shown to be very energy efficient and capable of impressive computational throughput when fully utilized by concurrent algorithms [18].”
BackgroundIMC: energy-efficient in-memory convolver for accelerating binarized deep neural network
Angizi, S., & Fan, D. (2017) · doi:10.1145/3183584.3183613 · Cited by 21
“It applies kernels (filters) by performing high-dimensional convolu-tions and then produces the features ( Output fmaps ) for the next layer[21, 22].”
Dyhard-DNN: even more DNN acceleration with dynamic hardware reconfiguration
Putic, M., Venkataramani, S., Eldridge, S., Buyuktosunoglu, A., Bose, P. K., & Stan, M. R. (2018) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1145/3195970.3196033 · Cited by 17
A cross-layer approach towards developing efficient embedded Deep Learning systems
Hanif, M. A., & Shafique, M. (2021) · Microprocessors and Microsystems · doi:10.1016/j.micpro.2020.103609 · Cited by 25
Trends and Opportunities for SRAM Based In-Memory and Near-Memory Computation
Srinivasa, S., Ramanathan, A. K., Sundaram, J., Kurian, D. J., Gopal, S., Jain, N., Srinivasan, A., Iyer, R., Narayanan, V., & Karnik, T. (2021) · International Symposium on Quality Electronic Design (ISQED) · doi:10.1109/isqed51717.2021.9424263 · Cited by 17
A Survey of Near-Data Processing Architectures for Neural Networks
Hassanpour, M., Riera, M., & González, A. M. (2022) · Machine Learning and Knowledge Extraction · doi:10.3390/make4010004 · Cited by 22
High-Throughput and Power-Efficient Convolutional Neural Network Using One-Pass Processing Elements
Sivasankari, B., Shunmugathammal, M., Appathurai, A., & Kavitha, M. (2022) · Journal of Circuits Systems and Computers · doi:10.1142/s0218126622502267 · Cited by 20
Xuan, Z., Liu, C., Zhang, Y., Li, Y., & Kang, Y. (2022) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2022.3224049 · Cited by 17
“Among them, general-purpose DNN accelerators represented by DaDianNao [2] and domain-specific CNN accelerators represented by Eyeriss [3] have been widely reported.”
MethodSAMBA: S parsity A ware In- M emory Computing B ased Machine Learning A ccelerator
Kim, D. E., Ankit, A., Wang, C., & Roy, K. (2023) · IEEE Transactions on Computers · doi:10.1109/tc.2023.3257513 · Cited by 17
Two-Level Scheduling Algorithms for Deep Neural Network Inference in Vehicular Networks
Wu, Y., Wu, J., Yao, M., Liu, B., Chen, L., & Lam, S.-K. (2023) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2023.3266795 · Cited by 18
Tsai, C.-W., Jiang, R., Zhang, L., Zhang, M., & Yoo, J. (2023) · IEEE Transactions on Biomedical Circuits and Systems · doi:10.1109/tbcas.2023.3327509 · Cited by 20
Efficient Processing of MLPerf Mobile Workloads Using Digital Compute-In-Memory Macros
Sun, X., Cao, W.-D., Crafton, B., Akarvardar, K., Mori, H., Fujiwara, H., Noguchi, H., Chih, Y., Chang, M.-F., Wang, Y., et al. (2024) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2023.3333290 · Cited by 15
Mirage: An RNS-Based Photonic Accelerator for DNN Training
Demirkiran, C., Yang, G., Bunandar, D., & Joshi, A. (2024) · Proceedings - International Symposium on Computer Architecture/Proceedings · doi:10.1109/isca59077.2024.00016 · Cited by 23
“We chose systolic arrays for their common usage in DNN acceleration and their superior performance over CPUs and GPUs [11], [31].”
BackgroundMethodStream: Design Space Exploration of Layer-Fused DNNs on Heterogeneous Dataflow Accelerators
Symons, A., Mei, L., Colleman, S., Houshmand, P., Karl, S., & Verhelst, M. (2024) · IEEE Transactions on Computers · doi:10.1109/tc.2024.3477938 · Cited by 40
“Dataflow accelerator (DA) architectures have evolved from single-core designs with specific dataflows [1], [2], [3] to multi-core and chiplet-based systems [4], [5], [6], [7], [8], increasing computing parallelism.”
Han, D., Lee, J., Lee, J., Choi, S., & Yoo, H. (2018) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas.2018.8351398 · Cited by 17
Hardware Accelerator Design for Machine Learning
Du, L., & Du, Y. (2018) · InTech eBooks · doi:10.5772/intechopen.72845 · Cited by 21
Shi, Y., Huang, Z., Oh, S., Kaslan, N., Song, J., & Kuzum, D. (2019) · IEEE Transactions on Electron Devices · doi:10.1109/ted.2019.2898402 · Cited by 20
Hardware implementation of RRAM based binarized neural networks
Huang, P. C., Zhou, Z., Zhang, Y., Xiang, Y., Han, R., Liu, L., Liu, X., & Kang, J. (2019) · APL Materials · doi:10.1063/1.5116863 · Cited by 18
Flow mapping and data distribution on mesh-based deep learning accelerator
Mirmahaleh, S. Y. H., Reshadi, M., Shabani, H., Guo, X., & Bagherzadeh, N. (2019) · IEEE/ACM International Symposium on Networks-on-Chip (NOCS) · doi:10.1145/3313231.3352378 · Cited by 21
“AlexNet flows were mapped on Maeri topology for five convolutional layers of AlexNet and Maeri had better latency and area consumption than Eyeriss, systolic array, and micro-switches array accelerators [12], [13], [16], [17].”
BackgroundMethodInfluentialFactored Radix-8 Systolic Array for Tensor Processing
Ullah, I., Inayat, K., Yang, J.-S., & Chung, J. (2020) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1109/dac18072.2020.9218585 · Cited by 18
“In [6], the authors target at accelerating the convolution operations in CNNs, study existing dataflows such as no local reuse, weight stationary, output stationary, and proposes a novel dataflow called row stationary.”
BackgroundInfluentialA Dynamic Timing Enhanced DNN Accelerator With Compute-Adaptive Elastic Clock Chain Technique
Jia, T., Ju, Y., & Gu, J. (2020) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2020.3027953 · Cited by 17
Yang, C., Hou, J., Wu, M., Mei, K., & Geng, L. (2020) · IEEE International Conference on Solid-State and Integrated Circuit Technology · doi:10.1109/icsict49897.2020.9278162 · Cited by 17
“Therefore, many reconfigurable CNN accelerators, such as Eyeriss [5], Thinker [6] and RNA [7], have been proposed to accelerate diverse CNN models.”
MethodFunctional Criticality Classification of Structural Faults in AI Accelerators
Chaudhuri, A., Talukdar, J., Su, F., & Chakrabarty, K. (2020) · IEEE International Test Conference (ITC) · doi:10.1109/itc44778.2020.9325272 · Cited by 17
“The ubiquitous application of DNNs has led to a rise in demand for custom artificial intelligence (AI) accelerators [1].”
BackgroundA High-speed Low-cost CNN Inference Accelerator for Depthwise Separable Convolution
Lin, Y., Li, R., He, W., Zho, X., He, J., Li, P., Jiang, Y., Liu, L., Wu, N., & Shi, C. (2020) · IEEE International Conference on Integrated Circuits, Technologies and Applications (ICTA) · doi:10.1109/icta50426.2020.9332057 · Cited by 17
“…[5]-[7], However, most of existing CNN accelerators are mainly developed for general large-scale CNN models and lack of exploiting the progress of model optimization [5], This problem has partially been addressed in [6] by proposing a hardware engine particularly for depthwise convolution.”
BackgroundA Lightweight Error-Resiliency Mechanism for Deep Neural Networks
Goldstein, B. F., Ferreira, V. C., Srinivasan, S., Das, D., Nery, A. S., Kundu, S. K., & França, F. M. G. (2021) · International Symposium on Quality Electronic Design (ISQED) · doi:10.1109/isqed51717.2021.9424287 · Cited by 23
“1: DNN system architecture composed by mesh of PE, each one with a multiply-and-accumulate (MAC) unit [7].”
BackgroundMethodNP-CGRA: Extending CGRAs for Efficient Processing of Light-weight Deep Neural Networks
Lee, J., & Lee, J. (2021) · Design, Automation, and Test in Europe · doi:10.23919/date51398.2021.9474256 · Cited by 17
PNeuro: A scalable energy-efficient programmable hardware accelerator for neural networks
Carbon, A., Philippe, J., Bichler, O., Schmit, R., Tain, B., Briand, D., Ventroux, N., Paindavoine, M., & Brousse, O. (2018) · Design, Automation, and Test in Europe · doi:10.23919/date.2018.8342165 · Cited by 21
Multi-dimensional Parallel Training of Winograd Layer on Memory-Centric Architecture
Hong, B., Ro, Y., & Kim, J. (2018) · IEEE/ACM International Symposium on Microarchitecture (MICRO) · doi:10.1109/micro.2018.00061 · Cited by 19
“To reduce the amount of communication for tile scattering, we use skipping of zero values [33], [34], and for tile gathering we propose an activation prediction method.”
BackgroundMethodAn Improved Low-Power Coding for Serial Network-On-Chip Links
Velayudham, S., Rajagopal, S., & Ko, S. (2019) · Circuits Systems and Signal Processing · doi:10.1007/s00034-019-01231-w · Cited by 17
Sabogal, S., George, A. D., & Crum, G. (2019) · IEEE Space Computing Conference (SCC) · doi:10.1109/spacecomp.2019.00010 · Cited by 27
“In [20], batch processing was employed to reuse convolution weights on multiple images to reduce memory accesses for improved bandwidth at the cost of increased latency for each image in the batch.”
BackgroundMethodMobile Machine Learning Hardware at ARM: A Systems-on-Chip (SoC) PerspectivePreprint
Zhu, Y., Mattina, M., & Whatmough, P. N. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1801.06274 · Cited by 23
“This is true even with a dedicated CNN hardware accelerator [12, 13, 15].”
MethodNeural Network-Hardware Co-design for Scalable RRAM-based BNN AcceleratorsPreprint
Kim, Y., Kim, H., & Kim, J. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1811.02187 · Cited by 26
“As a result, design of dedicated hardware for DNNs has been actively studied [1]–[3].”
BackgroundDNA: Differentiable Network-Accelerator Co-SearchPreprint
Zhang, Y., Fu, Y., Jiang, W., Li, C., You, H., Li, M., Chandra, V., & Lin, Y. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2010.14778 · Cited by 27
“Given a DNN and its acceleration specification, SOTA accelerators (Du et al., 2015; Chen et al., 2017; Zhao et al., 2020; Li et al., 2020a) explore different micro-architectures and algorithm-to-hardware mapping methods to maximize data reuses and thus acceleration efficiency.”
InfluentialHan, M., Wang, L., Xiao, L., Cai, T., Wang, Z., Xu, X., & Zhang, C. (2024) · IEEE Transactions on Computers · doi:10.1109/tc.2024.3398500 · Cited by 23
“In this manner, the input data, marked as the black lines, is transferred from the header PE in Sub-array A (PE[0, 0] and PE[1, 0]) to the tail PE in Sub-array D (PE[2, 0] and PE[2, 1]).”
BackgroundMethodDesign Methodology for Embedded Approximate Artificial Neural Networks
Balaji, A., Ullah, S., Das, A. K., & Kumar, A. S. (2019) · ACM Great Lakes Symposium on VLSI (GLSVLSI) · doi:10.1145/3299874.3319490 · Cited by 18
“Eyeriss [3] is a reconfigurable accelerator designed for CNNs that minimizes the energy cost for data movement.”
BackgroundSamal, K., Wolf, M., & Mukhopadhyay, S. (2020) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2020.3015889 · Cited by 21
Shatravin, V., Shashev, D., & Shidlovskiy, S. (2022) · Applied Sciences · doi:10.3390/app12105216 · Cited by 16
Shin, H., Kang, M., & Kim, L. (2022) · IEEE Transactions on Computers · doi:10.1109/tc.2022.3227871 · Cited by 16
“PROCESSING in-memory architectures [1], [2], [3] have been proposed to solve the memory bottleneck of the conventional DNN accelerators [4], [5], [6], [7].”
BackgroundBit-Line Computing for CNN Accelerators Co-Design in Edge AI Inference
Rios, M. A., Ponzina, F., Levisse, A., Ansaloni, G., & Atienza, D. (2023) · IEEE Transactions on Emerging Topics in Computing · doi:10.1109/tetc.2023.3237914 · Cited by 20
“This characteristic can, as we do in this papers, be effectively harnessed by Single Instruction, Multiple Data (SIMD) computation strategies to increase efficiency and performance [3].”
BackgroundVSPIM: SRAM Processing-in-Memory DNN Acceleration via Vector-Scalar Operations
Nie, C., Tang, C., Lin, J., Hu, H., Lv, C., Cao, T., Zhang, W., Jiang, L., Liang, X., Qian, W., et al. (2023) · IEEE Transactions on Computers · doi:10.1109/tc.2023.3285095 · Cited by 17
Fast and Scalable Multicore YOLOv3-Tiny Accelerator Using Input Stationary Systolic Architecture
Adiono, T., Ramadhan, R. M., Sutisna, N., Syafalni, I., Mulyawan, R., & Lin, C. H. (2023) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2023.3305937 · Cited by 19
“Numerous other state-of-the-art systolic array hardware implementations [12], [13], [14], [15], [16] include Google TPU [17], leveraged efficient intercore communication inside the accelerator, standardized PE layout for greater reconfigurability, and parallel data operation through dataflow…”
BackgroundSparse-DySta: Sparsity-Aware Dynamic and Static Scheduling for Sparse Multi-DNN Workloads
Fan, H., Venieris, S. I., Kouris, A., & Lane, N. D. (2023) · IEEE/ACM International Symposium on Microarchitecture (MICRO) · doi:10.1145/3613424.3614263 · Cited by 28
“Our scheduling assumes that the execution is performed in a per-layer or per-layer-block manner, which is a common setting in commercial AI hardware [30, 48] and existing research literature [4, 5, 15, 43, 52].”
BackgroundSundaram, S. S., Khodke, Y., Li, Y., Jang, S., Lee, S.-S., & Kang, M. (2023) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2023.3341348 · Cited by 17
“In addition, various compression techniques and supporting hardware have been introduced to reduce storage requirements and data movement costs [6], [20], [21].”
BackgroundEWS: An Energy-Efficient CNN Accelerator With Enhanced Weight Stationary Dataflow
Wang, C., Wang, Z., Li, S., Zhang, Y., Shen, H., & Huang, K. (2024) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2024.3359511 · Cited by 19
Giacomo, S. D., Ronchi, M., Borghi, G., Schaart, D. R., Carminati, M., & Fiorini, C. E. (2024) · IEEE Transactions on Radiation and Plasma Medical Sciences · doi:10.1109/trpms.2024.3378421 · Cited by 22
Liu, Y., Ma, Y., Zhang, B., Liu, L., Wang, J., & Tang, S. (2024) · Microelectronics Journal · doi:10.1016/j.mejo.2024.106197 · Cited by 22
Experimental demonstration of magnetic tunnel junction-based computational random-access memory
Lv, Y., Zink, B. R., Bloom, R. P., Cılasun, H., Khanal, P., Resch, S., Chowdhury, Z. I., Habiboglu, A., Wang, W., Sapatnekar, S. S., et al. (2024) · npj Unconventional Computing · doi:10.1038/s44335-024-00003-3 · Cited by 23
Liu, B., Huang, X., Zhang, Y., Yang, G., Yan, H., Zhang, C., Li, Z., Wang, Y., & Cai, H. (2024) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2024.3434706 · Cited by 17
“Eyeriss [9] proposed a row stationary dataflow that minimizes data access to high-cost storage in exchange for increased access to low-cost storage.”
BackgroundPipelining a triggered processing element
Repetti, T. J., Cerqueira, J. P., Kim, M. A., & Seok, M. (2017) · IEEE/ACM International Symposium on Microarchitecture (MICRO) · doi:10.1145/3123939.3124551 · Cited by 17
“Spatial accelerators support important workloads such as information retrieval [22], databases [33–35], string processing [26], and neural networks [8, 9].”
BackgroundA Quantized Training Method to Enhance Accuracy of ReRAM-based Neuromorphic Systems
Yang, Q. X., Li, H. H., & Wu, Q. (2018) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas.2018.8351327 · Cited by 16
“Eyeriss [3] focuses on the data flow and data reuse to avoid busy data transfer between computing engines, on-chip global buffer and off-chip DRAM chip.”
BackgroundA Continuous-Time Replication Strategy for Efficient Multicast in Asynchronous NoCs
Bhardwaj, K., & Nowick, S. M. (2018) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2018.2876856 · Cited by 16
“Finally, one-to-many and one-to-all patterns are also common in emerging NoC application areas, such as neuromorphic computing, both in spiking neural networks (SNNs) [14] as well as in deep convolutional neural networks [15].”
BackgroundChiu, P.-F., Choi, W. H., Ma, W., Qin, M., & Lueker-Boden, M. (2019) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas.2019.8702128 · Cited by 16
Convolutional Neural Network based for Automatic Text Summarization
Alquliti, W. H., & Ghani, N. B. A. (2019) · International Journal of Advanced Computer Science and Applications · doi:10.14569/ijacsa.2019.0100424 · Cited by 20
“The last layer is the classifier, which utilises these high-level features [34].”
BackgrounddMazeRunner: Optimizing Convolutions on Dataflow Accelerators
Dave, S., Shrivastava, A., Kim, Y., Avancha, S., & Lee, K. (2020) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp40776.2020.9054275 · Cited by 18
“So, it is crucial to automatically optimize the execution method for efficient acceleration of a convolution layer on a target dataflow accelerator and to explore efficient accelerator design for different convolution layers of multiple CNN models [5, 6].”
BackgroundMethodInfluentialA Case for Emerging Memories in DNN Accelerators
Mukherjee, A., Saurav, K., Nair, P., Shekhar, S., & Lis, M. (2021) · 2021 Design, Automation & Test in Europe Conference & Exhibition (DATE) · doi:10.23919/date51398.2021.9474252 · Cited by 12
“DNN accelerators [4, 6, 7] attempt to reduce the memory access energy through an on-chip memory hierarchy.”
BackgroundMethodOn Periodic Functions as Regularizers for Quantization of Neural Networks
Naumov, M., Diril, U., Park, J., Ray, B., Jablonski, J., & Tulloch, A. (2018) · arXiv · doi:10.48550/arxiv.1811.09862 · Cited by 12
A Quantized Neural Network Library for Proper Implementation of Hardware Emulation
Kiyama, M., Nakahara, Y., Amagasaki, M., & Iida, M. (2019) · 2019 Seventh International Symposium on Computing and Networking Workshops (CANDARW) · doi:10.1109/candarw.2019.00032 · Cited by 5
“In terms of the energy problem, energy-efficient DNN accelerators [18], [19] that support state-of-the-art CNNs have been developed.”
BackgroundDeploying human activity recognition in embedded RISC-V processors
Nunes, W. A., Reusch, R. S., Luza, L., Bernardon, E., Dal Zotto, A. E., Juracy, L. R., & Moraes, F. G. (2024) · Design Automation for Embedded Systems · doi:10.1007/s10617-024-09288-w
Metadata
- Authors
- Chen, Y.-H., Krishna, T., Emer, J. S., & Sze, V.
- Affiliations
- Massachusetts Institute of Technology (US)
- Georgia Institute of Technology (US)
- Nvidia (United States) (US)
- Venue
- IEEE Journal of Solid-State Circuits
- Published
- 2017-01
- Volume
- 52
- Issue
- 1
- Pages
- 127-138
- ISSN
- 0018-9200 (ISSN-L)
- 1558-173X
- OpenAlex
- W2289252105
- Semantic Scholar
- ffdaa12ef011de9dbf43be46d45a3abcc8288965
- Cited by
- 3,247
- References
- 42
- Citations by year
- 2026: 218 · 2025: 313 · 2024: 294 · 2023: 373 · 2022: 365 · 2021: 450 · 2020: 427 · 2019: 354 · 2018: 302 · 2017: 118
- FWCI
- 182.12
- Type
- journal article
- Language
- en
- Open access
- closed
- Locations
- IEEE Journal of Solid-State Circuits (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Hardware and Architecture › Parallel Computing and Optimization Techniques
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Advanced Memory and Neural Computing
- Keywords
- reconfigurable accelerators, deep convolutional neural networks, energy efficiency, spatial architecture, data movement reduction, data compression, processing elements, AlexNet, VGG16, convolutional layers
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.