CONV-SRAM: An Energy-Efficient SRAM With In-Memory Dot-Product Computation for Low-Power Convolutional Neural Networks
Open accessThe DOI matches its Crossref record; 25 of the 29 references it cites were found in a registry or an index.
Abstract
This paper presents an energy-efficient static random access memory (SRAM) with embedded dot-product computation capability, for binary-weight convolutional neural networks. A 10T bit-cell-based SRAM array is used to store the 1-b filter weights. The array implements dot-product as a weighted average of the bitline voltages, which are proportional to the digital input values. Local integrating analog-to-digital converters compute the digital convolution outputs, corresponding to each filter. We have successfully demonstrated functionality (>98% accuracy) with the 10 000 test images in the MNIST hand-written digit recognition data set, using 6-b inputs/outputs. Compared to conventional full-digital implementations using small bitwidths, we achieve similar or better energy efficiency, by reducing data transfer, due to the highly parallel in-memory analog computations.
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
“Among charge-based memories, SRAM cells (a bi-stable transistor structure typically made of two CMOS inverters connected back-to-back which retains a charge concentration, see Fig. 3a for an example of the structure of a crossbar array of 6T SRAM) have been widely used for VMM75–77.”
From Hardware implementation of memristor-based artificial neural networks (2024) · cites this paper as [76]
Citations
Cites
ImageNet 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 · cited 1×
“Convolutional neural networks (CNNs) provide state-of-theart results in a wide variety of AI/ML applications, ranging from image classification [3] to speech recognition [1].”
BackgroundGradient-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 5×
“However, for the MNIST dataset with LeNet-5 CNN, we estimate [22] those to have only small contributions to the overall energy-efficiency per MAV operation, due to the high parallelism supported by our in-memory approach.”
MethodInfluentialHinton, G., Deng, L., Yu, D., Dahl, G., Mohamed, A.-R., Jaitly, N., Senior, A., Vanhoucke, V., Nguyen, P., Sainath, T., et al. (2012) · IEEE Signal Processing Magazine · doi:10.1109/msp.2012.2205597 · Cited by 10,260
DeepFace: Closing the Gap to Human-Level Performance in Face Verification
Taigman, Y., Yang, M., Ranzato, M., & Wolf, L. (2014) · 2014 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2014.220 · Cited by 6,542 · cited 1×
“Facial recognition [2] enables using our faces to get access to devices in a more intuitive manner, instead of traditional passcodes.”
BackgroundXNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks
Rastegari, M., Ordonez, V., Redmon, J., & Farhadi, A. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46493-0_32 · Cited by 4,408 · cited 2×
“[6] proposed a binary-weight-network (BWN), where the filter weights (wi s) can be trained to be +1/−1 (with a common scaling factor per 3-D filter: α).”
BackgroundMethodEfficient Processing of Deep Neural Networks: A Tutorial and Survey
Sze, V., Chen, Y.-H., Yang, T.-J., & Emer, J. S. (2017) · Proceedings of the IEEE · doi:10.1109/jproc.2017.2761740 · Cited by 4,215 · cited 1×
“In the conventional all-digital implementation of CNNs [4], [5], [9], with the memory and the processing elements being physically separate, reading wi s and the partial sums from the on-chip SRAMs lead to a lot of data movement per computation [10] and, hence, make them energy-hungry.”
BackgroundEyeriss: An Energy-Efficient Reconfigurable Accelerator for Deep Convolutional Neural Networks
Chen, Y.-H., Krishna, T., Emer, J. S., & Sze, V. (2017) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2016.2616357 · Cited by 3,247 · cited 12×
“CNNs typically consist of a cascade of convolutional (CONV) and fully-connected (FC) layers (Fig.”
BackgroundMethodInfluentialBinaryConnect: Training Deep Neural Networks with binary weights during propagations
Courbariaux, M., Bengio, Y., & David, J.-P. (2015) · arXiv · doi:10.48550/arxiv.1511.00363 · Cited by 2,180 · cited 1×
“BWN performs better than binary connect [7], which does not incorporate the scaling factor of α per filter, and also binarized neural networks [8], where both weights and activations are constrained to ±1.”
Background1.1 Computing's energy problem (and what we can do about it)
Horowitz, M. (2014) · 2014 IEEE International Solid-State Circuits Conference Digest of Technical Papers (ISSCC) · doi:10.1109/isscc.2014.6757323 · Cited by 2,069 · cited 1×
“to access data from memory can be much higher than the energy needed for a compute operation with that data [11].”
BackgroundHubara, I., Soudry, D., & Yaniv, R. E. (2016) · arXiv · doi:10.48550/arxiv.1602.02505 · Cited by 1,667 · cited 1×
“BWN performs better than binary connect [7], which does not incorporate the scaling factor of α per filter, and also binarized neural networks [8], where both weights and activations are constrained to ±1.”
Background
Research works printed but not linked: 1
- whatmough, 2017, A 28 nm SoC with a 1.2 GHz 568 nJ/prediction sparse deep-neural-network engine with >0.1 timing error rate tolerance for IoT applications, IEEE Int Solid-State Circuits Conf (ISSCC) Dig Tech Papers, 242
Cited by
Items 201–300 of 414
Lee, K., Kim, J., & Park, J. (2022) · doi:10.1145/3508352.3549423 · Cited by 6
A Digital Bit-Reconfigurable Versatile Compute-In-Memory Macro for Machine Learning Acceleration
Zhang, X., Lu, Y., Wang, B., & Kim, T. T.-H. (2023) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2023.3257058 · Cited by 6
“In [10], [11], the authors proposed analog-based CIM macros only for CNN acceleration.”
BackgroundMethodAn 8-T Processing-in-Memory SRAM Cell-Based Pixel-Parallel Array Processor for Vision Chips
Chen, L., Shi, C., He, J., Yu, J., Wang, H., Lu, J., Liu, L., Wu, N., & Tian, M. (2023) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2023.3311206 · Cited by 8
“The state-of-the-art SRAM based PIM works mainly adopt 6T [32], [35], [37], 7T [43], 8T [33], [34], [36], [39], and 10T [30] SRAM to implement processing-in-memory.”
MethodLee, Y. K., Ko, D. H., Cho, S., Yeo, M., Kang, M., & Jung, S. (2023) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2023.3337754 · Cited by 6
“For multi-bit input and binary weight operation, BL DAC and large area 10T SRAM are used in [8], and WL DAC and 6T SRAM are used in [9].”
MethodBai, Y., Li, Y., Zhang, H., Jiang, A., Du, Y., & Du, L. (2024) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2024.3366025 · Cited by 9
“The parameter S denotes the fractional length and a larger S indicates higher-quantization resolution ( The value of S can be calculated by (3), where B q denotes quantization bit-width, which is set by users.”
BackgroundGu, Z., Dou, S., You, H., Zhan, Y., Qiao, S., & Zhou, Y. (2024) · IEEE Access · doi:10.1109/access.2024.3369057 · Cited by 6
“Over the last few years, SRAM-based CIM designs have demonstrated their advantage in significantly improving energy efficiency through reducing data movement, showing potential application in deep learning (DL) based edge computing devices [1], [2], [3], [4], [5], [6], [7], [8].”
BackgroundMethodConfigurable in-memory computing architecture based on dual-port SRAM
Zhao, Y., Liu, Y., Zheng, J., Tong, Z., Wang, X., Yu, R., Wu, X., Zhou, Y., Peng, C., Lu, W., et al. (2024) · Microelectronics Journal · doi:10.1016/j.mejo.2024.106163 · Cited by 6
A PWM-DAC for Analog In-Memory Computing in Mixed-Signal Accelerators
Malena, F., Boni, A., & Caselli, M. (2023) · Conference on Ph.D. Research in Microelectronics and Electronics · doi:10.1109/prime58259.2023.10161781 · Cited by 7
“as fast decision, less communication traffic, and higher data security [2].”
BackgroundLi, J., Li, K., Un, K.-F., Yu, W.-H., Martins, R. P., & Mak, P. (2025) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2025.3532544 · Cited by 8
NS-Engine: Near-Sensor Neural Network Engine with SRAM-Based Compute-in-Memory Macro
Ren, E., Liu, J., Luo, L. Z., Yang, X., Wei, Q., & Qiao, F. (2024) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas58744.2024.10557977 · Cited by 5
Low-Power 8T SRAM Compute-in-Memory Macro for Edge AI Processors
Shin, H.-J., & Jo, S.-H. (2024) · Applied Sciences · doi:10.3390/app142310924 · Cited by 5
“By replacing traditional Von Neumann architectures, CIM offers effective solutions to various optimization and mathematical problems, enabling AI edge devices to perform more efficiently within their constrained environments and meet the demands of real-time, data-intensive applications [8].”
Kim, J., Lee, K., & Park, J. (2022) · International Symposium on Low Power Electronics and Design (ISLPED) · doi:10.1145/3531437.3539718 · Cited by 5
Wang, X., Liu, X., Hu, X., Zhong, X., Chen, X., Liu, Y., Kong, P., Tian, F., & Tsui, C.-Y. (2022) · IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas54282.2022.9869970 · Cited by 5
“SRAM-CIM incorporates multiply-and-accumulate (MAC) operation into the bit cells of an SRAM array and is configurable to support a wide range of neural networks, including binarized neural networks (BNN) which are designed to reduce the size of weight stored in the memory.”
BackgroundMethodInfluentialSeo, S. H., Kim, S., Jung, S. J., Kwon, S., Lee, H., & Lee, J. W. (2022) · European Solid-State Circuits Conference (ESSCIRC) · doi:10.1109/esscirc55480.2022.9911343 · Cited by 7
“a) The effect of computations outside the self-attention layers are excluded as explained in Section IV b) Die area is approximated with the number of shader cores c) 90% output sparsity d) 88% output sparsity e) Assumes 40nm implementation and energy is proportional to (Technology)2 [14]”
Method4b/4b/8b Precision Charge-Domain 8T-SRAM Based CiM for CNN Processing
Zang, Q., Goh, W. L., Chong, Y. S., & Do, A. T. (2023) · IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas57966.2023.10168593 · Cited by 6
“w[0] w[l] w[2] w[3] :C - -C —j— C =}=2C =}=2C”
BackgroundMethodResultInfluentialDSPIMM: A Fully Digital SParse In-Memory Matrix Vector Multiplier for Communication Applications
Sridharan, A., Zhang, F., Sui, Y., Yuan, B., & Fan, D. (2023) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1109/dac56929.2023.10247829 · Cited by 5
“2144751 can be effectively addressed via in-memory computing (IMC), a technique that has been well-studied to develop low-power general NN hardware [8]–[11].”
BackgroundJiang, L., Sun, S., Ge, J., Zhang, H., & Kang, W. (2023) · IEEE Non-Volatile Memory Systems and Applications Symposium (NVMSA) · doi:10.1109/nvmsa58981.2023.00026 · Cited by 5
SRAM-Based Hybrid Analog Compute-In-memory Architecture to Enhance the Signal Margin
Kushwaha, D., Joshi, R., Dasgupta, S., & Bulusu, A. (2024) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas58744.2024.10558054 · Cited by 5
“The 4-bit input activation is first converted into an equivalent analog voltage (V a ) through digital to analog converter [2] and then used to charge the read bitlines (RBLs) of the 10T SRAM bit-cell for the multiplication operation.”
MethodInfluentialFlash Based In-Memory Multiply-Accumulate Realisation: A Theoretical Study
Balagopal, S. A., & Viraraghavan, J. (2020) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas45731.2020.9180925 · Cited by 5
Modelling and Optimization of a Mixed-Signal Accelerator for Deep Neural Networks
Caselli, M., & Boni, A. (2023) · International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design · doi:10.1109/smacd58065.2023.10192111 · Cited by 5
“Extending machine intelligence to edge devices enables fast decisions, limits the communication with the cloud, and improves data security [2].”
BackgroundChang, L., Zhao, X., Yue, T., Yang, X., Li, C., Lin, S., & Zhou, J. (2023) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2023.3330648 · Cited by 6
“Biswas and Chandrakasan [25] presented an energy-efficient SRAM-based CIM macro with 6-bit inputs/output, which achieves 40.3-TOPS/W energy efficiency.”
BackgroundInfluentialCross-coupled 4T2R multi-logic in-memory computing circuit design
Lin, Z., Yue, C., Li, K., Feng, Q., Li, S., Zhao, Y., Wang, Y., Chen, J., Lu, W., Peng, C., et al. (2024) · Microelectronics Journal · doi:10.1016/j.mejo.2024.106179 · Cited by 5
Kim, H., An, Y., Son, D., Kim, J. J., & Shim, Y. S. (2025) · IEEE transactions on circuits and systems for artificial intelligence. · doi:10.1109/tcasai.2025.3592865 · Cited by 4
“The energy efficiency (J/SOP) decreases according to the following equation [42]: Energy ∝ Tech 2 .”
Joseph, B., & Kavitha, R. K. (2025) · AEU - International Journal of Electronics and Communications · doi:10.1016/j.aeue.2025.156017 · Cited by 4
Chen, Y., Yin, G., Lee, M., Tang, W., Yang, Z., Liu, Y., Yang, H., & Li, X. (2022) · International Conference on Computer Aided Design · doi:10.1145/3508352.3549335 · Cited by 4
A Computing In-Memory Multibit Multiplication Based on Decoupling and In-Array Storing
Zhang, J., Tong, Z., Wang, H. E., Wang, X., Zhao, Q., Zhou, J., Wang, J., Lin, Z., & Wu, X. (2024) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2024.3364715 · Cited by 4
“Currently, two main types of multibit multiplication mapping circuits exist in CIM: analog-domain [4] [8], [9], [10], [11], [12], [13] and digital-domain [14], [15], [16], [17].”
BackgroundLinearity Performance of Charge Domain In-Memory Computing: Analysis and Calibration
Zhang, H., Bai, Y., Shen, J., Du, Y., & Du, L. (2024) · Integrated Circuits and Systems · doi:10.23919/ics.2024.3422968 · Cited by 5
SRAM-Based Processing-in-Memory (PIM)
Kim, H., Yu, C., & Kim, B. (2022) · Springer eBooks · doi:10.1007/978-3-030-98781-7_3 · Cited by 4
Linearity Analysis for Charge Domain In-memory Computing
Zhang, H., Du, Y., & Du, L. (2023) · IEEE International Conference on ASIC (ASICON) · doi:10.1109/asicon58565.2023.10396087 · Cited by 4
“Compute-in-memory (CIM) architectures offer a solution to reduce power consumption in data movement by integrating arithmetic units directly into the memory [4].”
BackgroundKhosravi, H., Carmona‐Galán, R., Fernández‐Berni, J., & Kandalaft, N. (2024) · IEEE Annual Computing and Communication Workshop and Conference (CCWC) · doi:10.1109/ccwc60891.2024.10427792 · Cited by 4
“Additionally, during the writing operation, the suggested circuit demonstrates improved write SNM in comparison to both the conventional 6T and 10T SRAM bit-cells [18].”
MethodResultInfluentialKushwaha, D., Joshi, A., Goel, A., Joshi, R., Dasgupta, S., & Bulusu, A. (2024) · International Symposium on Quality Electronic Design (ISQED) · doi:10.1109/isqed60706.2024.10528768 · Cited by 4
“The MOSFET's V DS non-idealities disturb the MAC operation's linearity and reduce the available voltage swing on bitlines for the MAC [13], degrading the signal margin for robust MAC operation.”
MethodA PWM-SAR ADC with Asynchronous Control Unit for Mixed-Signal Accelerators
Caselli, M., Boni, A., & Caselli, S. (2024) · International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design · doi:10.1109/smacd61181.2024.10745459 · Cited by 4
Time to Learn: Temporal Accelerators as an Embedded Deep Neural Network Platform
Cichiwskyj, C., Qian, C., & Schiele, G. (2020) · Communications in computer and information science · doi:10.1007/978-3-030-66770-2_19 · Cited by 4
A 2D2R ReRAM CIM accelerator for multilayer perceptron in visual classification applications
Zhou, K., Lin, R., Guo, Z., Liu, Y., Jiang, J., Zhao, C., Fang, J., Xu, X., Xue, X., & Zeng, X. (2022) · Microelectronics Journal · doi:10.1016/j.mejo.2022.105478 · Cited by 4
ReRAM-Based Processing-in-Memory (PIM)
Kim, T. T.-H., Lu, L., & Chen, Y. (2022) · Springer eBooks · doi:10.1007/978-3-030-98781-7_5 · Cited by 4
Chen, Y., Fu, Y., Lee, M., George, S., Liu, Y., Narayanan, V., Yang, H., & Li, X. (2022) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2022.3231589 · Cited by 6
“It is noted that there are some emerging SRAM-based CIM schemes to reduce the data transfer costs by enabling inSRAM computing [10], [11], [12].”
BackgroundLow Power Ternary XNOR using 10T SRAM for In-Memory Computing
Lee, S., & Kim, Y. (2022) · International SoC Design Conference · doi:10.1109/isocc56007.2022.10031490 · Cited by 4
“The advantage of IMC is that because it computes in memory, the amount of data moving between memory and processor is significantly smaller [2].”
BackgroundZhang, X., Sharma, V., Lu, Y., Jo, Y.-J., & Kim, T. T.-H. (2023) · IEEE Asian Solid-State Circuits Conference (A-SSCC) · doi:10.1109/a-sscc58667.2023.10347952 · Cited by 4
A 28-nm 9T SRAM-based CIM macro with capacitance weighting module and redundant array-assisted ADC
Lin, Z., Yu, R., Huo, D., Zhu, Q., Long, M., Qin, Y., Liu, Y., Zhao, Y., Chen, L., Wang, T., et al. (2024) · Microelectronics Journal · doi:10.1016/j.mejo.2024.106397 · Cited by 4
Kernel approximation using analogue in-memory computing
Büchel, J., Camposampiero, G., Vasilopoulos, A., Lammie, C., Gallo, M. L., Rahimi, A., & Sebastian, A. (2024) · Nature Machine Intelligence · doi:10.1038/s42256-024-00943-2 · Cited by 8
Custom Memory Design for Logic-in-Memory: Drawbacks and Improvements over Conventional Memories
Ottati, F., Turvani, G., Masera, G., & Vacca, M. (2021) · Electronics · doi:10.3390/electronics10182291 · Cited by 4
Chang, S.-H., Yen, R.-H., & Liu, C. J. (2025) · ACM Journal on Emerging Technologies in Computing Systems · doi:10.1145/3711830 · Cited by 3
“Therefore, Computing In-Memory (CIM) [1] is proposed to solve the issues as shown in Figure 1(b).”
Xue, C., Yang, Y., He, S., Du, G., Wang, Y., & He, Y. (2025) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2025.3582263 · Cited by 3
“The majority of contemporary CIM research endeavors have primarily focused on optimizing integer MAC operations [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24].”
Venu, B., Rao, K. T., Penumalli, K., Yellampalli, S., & Vaddi, R. (2024) · Semiconductor Science and Technology · doi:10.1088/1361-6641/ad3273 · Cited by 5
“The authors in [23] have been presented the convolution SRAM for a machine learning-based dot product system.”
Han, X., Zhao, R., Liu, Y., Liu, Y., & Chen, J. (2024) · Micromachines · doi:10.3390/mi15081056 · Cited by 4
“IMC based on SRAM primarily involves operations such as multiply–accumulate (MAC) [9,10], multi-bit multiplication [11,12], and Boolean logic, with a key focus on logic-in-memory operations in this paper.”
BackgroundKim, T., & Chung, Y. (2024) · Applied Sciences · doi:10.3390/app14219749 · Cited by 3
Lee, B., Ji, W., Kim, H., Han, S., Park, G., Hur, P., Jeon, G., Lee, H., Chung, Y., Son, J., et al. (2024) · Advanced Intelligent Systems · doi:10.1002/aisy.202400600 · Cited by 3
Chen, Y., Du, X., Yin, G., Tang, W., Liu, Y., Yang, H., & Li, X. (2025) · Asia and South Pacific Design Automation Conference (ASP-DAC) · doi:10.1145/3658617.3697570 · Cited by 3
“SRAM-based compute-in-memory (SRAM-CiM) [6][7] has been proposed in recent years as a promising technique to address the memory wall.”
Jiang, W., Caron, C., Avasare, P., Pauwels, M., Verhelst, M., & Dehaene, W. (2025) · IEEE Journal of Solid-State Circuits · doi:10.1109/jssc.2025.3557967 · Cited by 3
“I N-MEMORY-COMPUTE (IMC) is gaining increasing attention from both academia and industry because of its potential for higher energy efficiency, while alleviating the memory wall bottleneck [1], [2], [3], [4], [5], [6], [7], [8], [9], [10].”
Dual-Mode 2T1C DRAM Process-In-Memory Architecture for Boolean and MAC Operations
An, Y., Kim, H., Son, D., Yu, H., & Shim, Y. S. (2025) · IEEE Access · doi:10.1109/access.2025.3592352 · Cited by 3
“Meanwhile, various analog circuit techniques are used to perform arithmetic operations for the analog SRAM PIM system [5], [6], [7].”
Elastic AI: system support for adaptive machine learning in pervasive computing systems
Cichiwskyj, C., Schmeißer, S., Qian, C., Einhaus, L., Ringhofer, C., & Schiele, G. (2021) · CCF Transactions on Pervasive Computing and Interaction · doi:10.1007/s42486-021-00070-6 · Cited by 3
“Other approaches aim to develop special ASICs for neural processing units (NPUs), e.g. Google’s tensor processor unit (Jouppi et al. 2018), or CONV-SRAM (Biswas and Chandrakasan 2018).”
BackgroundA High Accuracy Low Power Convolution Operator with 12T SRAM for CNN
Oh, T. S., Pu, Y., & Lee, K. (2021) · International Conference on Ubiquitous and Future Networks · doi:10.1109/icufn49451.2021.9528655 · Cited by 3
“The circuit of convolution is implemented in [3].”
BackgroundMethodIn‐memory calculation with embedded arithmetic and logic units for deep neural network
Yu, S., Fu, J., Lin, Z., Peng, C., & Wu, X. (2022) · Electronics Letters · doi:10.1049/ell2.12549 · Cited by 3
Design and Challenges of Edge Computing ASICs on Front-End Electronics
Miryala, S., Carini, G. A., Deptuch, G. W., Huang, J., Katkoori, S., Maj, P. G., Mandal, S., Ren, Y., & Zaman, M. A. (2022) · International Symposium on Quality Electronic Design (ISQED) · doi:10.1109/isqed54688.2022.9806248 · Cited by 4
Systolic-RAM: Scalable Direct Convolution Using In-Memory Data Movement
Rohan, J. N., & Kulkarni, J. P. (2022) · IEEE Solid-State Circuits Letters · doi:10.1109/lssc.2022.3173560 · Cited by 3
“Duplication occurs since the indirect convolution method was tailor-made for vector-matrix multiplication (VMM)-based accelerators and uses an image-to-column (IM2COL) transformation [3]–[5].”
MethodYang, K., Wang, Z., Zheng, Y., Zheng, P., Yao, S., Zhu, X., Yue, Y., Wang, W., Zhang, J., & Ma, J. (2022) · Remote Sensing · doi:10.3390/rs14215315 · Cited by 3
Bipolar flash for bifunctional computing operations
Kiani, F., & Xia, Q. (2023) · Nature Nanotechnology · doi:10.1038/s41565-023-01358-7 · Cited by 3
Iqbal, B., Grover, A., & Rawat, H. (2023) · International Conference on VLSI Design · doi:10.1109/vlsid57277.2023.00037 · Cited by 3
“Depending on the full-scale range (FSR) of the bitlines and MAC value distribution, [6] has chosen 11 levels ADC while [7] employed 32 levels ADC.”
MethodAGNI: In-Situ, Iso-Latency Stochastic-to-Binary Number Conversion for In-DRAM Deep Learning
Shivanandamurthy, S. M., Vatsavai, S. S., Thakkar, I. G., & Salehi, S. A. (2023) · International Symposium on Quality Electronic Design (ISQED) · doi:10.1109/isqed57927.2023.10129301 · Cited by 3
Analysis of resistive defects on a foundry 8T SRAM-based IMC architecture
Ammoura, L., Flottes, M.-L., Girard, P., Noël, J.-P., & Virazel, A. (2023) · Microelectronics Reliability · doi:10.1016/j.microrel.2023.115029 · Cited by 3
GRAPHIC: Gather and Process Harmoniously in the Cache With High Parallelism and Flexibility
Chen, Y., Lee, M., Dai, G., Zhou, M., Challapalle, N. R., Wang, T., Yu, Y., Liu, Y., Wang, Y., Yang, H., et al. (2023) · IEEE Transactions on Emerging Topics in Computing · doi:10.1109/tetc.2023.3290683 · Cited by 3
“The outputs are usually obtained by comparators [14], ADCs [15], or adder-trees [16], [17], which can deal with dozens (or even far more) rows concurrently.”
BackgroundUnlocking the resistive switching in Acacia Senegal-based electrolyte for neuromorphic computation
Lokhandwala, A., Thakkar, P., Gosai, J., Oza, S., & Solanki, A. (2024) · Journal of Materials Chemistry C · doi:10.1039/d3tc03451h · Cited by 3
An in-Array Build-In Self-Test Scheme for Embedded SRAM Array
Wei, F., Cui, X., & Zhang, S. (2024) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2024.3375637 · Cited by 3
“Biswas and Chandrakasan [11] implemented the dot-production for the convolutional neural networks (CNNs) based on a 10T SRAM array.”
MethodSaragada, P. K., & Das, B. P. (2024) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2024.3422883 · Cited by 3
“On the other hand, the SRAM-based in-memory computation (IMC) architecture [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20] eliminates the above-mentioned problems by performing the MAC operations directly (near) at the place where the…”
BackgroundMethodMa, S., Mhatre, K. M., Weng, J., Hanindhito, B., Wang, Z., Nowatzki, T., John, L. K., & Arora, A. (2024) · ACM Transactions on Architecture and Code Optimization · doi:10.1145/3690824 · Cited by 6
“If not, we invoke a feedback loop to explore a more conservative code organization for less memory occupancy.”
Liu, Y. H., Wang, J., Liu, S., Sun, M., Zhang, X., Zhou, J., Yan, S., Pan, R., Hu, H., & Liu, Y. (2024) · Microelectronics Journal · doi:10.1016/j.mejo.2024.106470 · Cited by 3
Ghazal, O. G., Wang, W. J., Kvatinsky, S., Merchant, F., Yakovlev, A., & Shafik, R. (2025) · Philosophical Transactions of the Royal Society A Mathematical Physical and Engineering Sciences · doi:10.1098/rsta.2023.0393 · Cited by 4
8T-SRAM Computing-in-Memory Macro with Bitline Leakage Compensation
Peng, C., Chen, J., Zhao, M., Lu, W., Lin, Z., Hao, L., Li, X., Dai, C., & Wu, X. (2025) · Circuits Systems and Signal Processing · doi:10.1007/s00034-025-03052-6 · Cited by 3
Chen, Q., Su, Y., Kim, H., Yoo, T., Kim, T. T.-H., & Kim, B. (2020) · Design, Automation, and Test in Europe · doi:10.23919/date48585.2020.9116349 · Cited by 3
Place recognition with deep superpixel features for brain-inspired navigation
Zhao, J. H., Tang, J., Zhao, D., Cao, H., Liu, X., Shen, C., Wang, C., & Liu, J. (2020) · Review of Scientific Instruments · doi:10.1063/5.0027767 · Cited by 3
CIM-Based Smart Pose Detection Sensors
Chou, J., Chang, T.-W., Liu, X.-Y., Wu, T., Chen, Y., Hsu, Y., Chen, C., Liu, T.-T., & Shih, C. (2022) · Sensors · doi:10.3390/s22093491 · Cited by 3
Kim, B. (2019) · IEEE International System-on-Chip Conference (SOCC) · doi:10.1109/socc46988.2019.1570571922 · Cited by 3
“Instead of utilizing the standard SRAM bitcell as weight storage and a computation unit, a different type of SRAM in-memory computing macros [8-9] based on voltage-domain computations used SRAM cells only for storing weights.”
MethodA 22-nm FDSOI 8T SRAM Based Time-Domain CIM for Energy-Efficient DNN Accelerators
Zhou, Y., Cheng, Z., Liu, H., Xiong, T., & Wang, B. (2022) · IEEE Asia Pacific Conference on Circuits and Systems (APCCAS) · doi:10.1109/apccas55924.2022.10090315 · Cited by 3
“Previous CIM works utilized currentdomain or voltage-domain analog computation, which can operate MAC effectively, but finite analog headroom and additional circuits limiting the processing of high-precision data and area efficiency [6-8].”
MethodMitigating Power and Process Variation for Analog CIM Design Migration
Chang, S.-H., Chen, S.-Y., Yang, C., Huang, H.-W., Yang, Y.-C., Chen, W., Liu, C. J., & Hung-Ming, C. (2024) · International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design · doi:10.1109/smacd61181.2024.10745463 · Cited by 2
“To handle this issue, computing-in-memory (CIM) technique [1] has been proposed to avoid data movement via computing convolutions in the memory macro.”
MethodZhang, X., Cheng, Y., Zou, D., Gu, K., Wang, M., & Wang, Z. (2025) · Journal of Systems Architecture · doi:10.1016/j.sysarc.2025.103512 · Cited by 3
Wu, H., & Lin, Z. (2025) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2025.3613056 · Cited by 2
“Thus, computation-in-memory (CIM) integrates computation within memory arrays to alleviate the Von-Neumann bottleneck [16], [17], [18], [19], [20].”
RAPID : R e-Configurable A dder Tree Based P erformance- I ncentivized AI D igital CIM Macro
Sankhe, A., Lokhande, M., Dhakad, N. S., Sharma, R., & Vishvakarma, S. K. (2026) · IEEE Transactions on Nanotechnology · doi:10.1109/tnano.2026.3680063 · Cited by 2
“5(c), which involves evaluation for LeNet-5/MNIST, VGG-like/Custom dataset, AlexNet/CIFAR-10 and Custom-8/A-Z Handwritten dataset at 1A4W, 2A4W, 4A4W and 8A8W and is compared with FP32-tensor and SoTA works [6], [7], [9], [14].”
InfluentialKim, Y., Jin, H., Kim, D., Ha, P.-B., Park, M., Hwang, J., Lee, J., Woo, J.-M., Choi, J., Lee, C., et al. (2023) · Electronics · doi:10.3390/electronics12030678 · Cited by 3
Arun, A., V, Hareesh, V., A, A. N. S., & Sajeesh, M. (2024) · International Conference on Trends in Engineering Systems and Technologies (ICTEST) · doi:10.1109/ictest60614.2024.10576118 · Cited by 2
Dee, A., Vuong, D., Bennett, K., & Moazeni, S. (2024) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2024.3422809 · Cited by 3
“Recently various SRAM-based CIM realizations have been proposed for matrix multiplication acceleration in neural networks [20], [21], [22].”
Energy-Efficient Time-Domain Computation for Edge Devices: Challenges and Prospects
Maharmeh, H. A., Ismail, M., & Alhawari, M. (2024) · Foundations and Trends® in Integrated Circuits and Systems · doi:10.1561/3500000013 · Cited by 3
An Incremental Time-Domain Mixed-Signal Matrix-Vector-Multiplication Technique for Low-Power Edge-AI
Hérissé, K., Larras, B., Stefanelli, B., Kaiser, A., & Frappé, A. (2024) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2024.3480154 · Cited by 2
LUNA-CiM: A Programmable Compute-in-Memory Fabric for Neural Network Acceleration
Dehghanzadeh, P., Sen, O., Chatterjee, B., & Bhunia, S. (2025) · IEEE Transactions on Computers · doi:10.1109/tc.2025.3525601 · Cited by 2
“[6], [7].”
Be CIM or Be Memory: A Dual-mode-aware DNN Compiler for CIM Accelerators
Zhao, S., Li, Y., Li, B., He, Y., Wang, M., Han, Y., & Wang, Y. (2025) · International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS) · doi:10.1145/3676641.3716248 · Cited by 2
“This mode-switching functionality is controlled by modifying the input signals on the global lines.”
Dai, C., Zhang, J., Wang, R., Chen, J., Hu, W. S., Hao, L., Lu, W., Lin, Z., Peng, C., & Wu, X. (2025) · Microelectronics Journal · doi:10.1016/j.mejo.2025.106703 · Cited by 2
Convergent Time-Stepping Schemes for Analog ReLU Networks
Elfadel, I. M. (2021) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas51556.2021.9401124 · Cited by 2
Monga, K., Shenoy, M. V., Chaturvedi, N., & Gurunarayanan, S. (2023) · Analog Integrated Circuits and Signal Processing · doi:10.1007/s10470-023-02169-5 · Cited by 2
In-memory Computing based Boolean and logical Circuit Design using 8T SRAM
Yeswanth, C., & Acharya, A. (2021) · Devices for Integrated Circuit (DevIC) · doi:10.1109/devic50843.2021.9455869 · Cited by 2
A Reliability-Concerned Compute-in-Memory Behavior Model for Convolutional Neural Network
Cheng, K., Song, J., Zhang, X., He, Y., Wang, R., & Wang, Y. (2021) · IEEE International Symposium on the Physical and Failure Analysis of Integrated Circuits (IPFA) · doi:10.1109/ipfa53173.2021.9617344 · Cited by 3
Gu, Z., Zhao, H., Wang, X., Qiao, S., & Zhou, Y. (2022) · Electronics · doi:10.3390/electronics11091376 · Cited by 2
Hadamard product-based in-memory computing design for floating point neural network training
Fan, A., Fu, Y., Tao, Y., Jin, Z., Han, H., Liu, H., Zhang, Y., Yan, B., Yang, Y., & Huang, R. (2023) · Neuromorphic Computing and Engineering · doi:10.1088/2634-4386/acbab9 · Cited by 3
SiCBit-PUF: Strong in-Cache Bitflip PUF Computation for Trusted SoCs
Xynos, A., Tenentes, V., & Tsiatouhas, Y. E. (2023) · IEEE European Test Symposium (ETS) · doi:10.1109/ets56758.2023.10173941 · Cited by 2
Towards Model-Size Agnostic, Compute-Free, Memorization-based Inference of Deep Learning
Giacomini, D., Hashem, M. B., Suarez, J., & Trivedi, A. R. (2024) · International Conference on VLSI Design · doi:10.1109/vlsid60093.2024.00036 · Cited by 2
“MBI requires higher energy than one-bit weight CIM design [14]; however, the applicability of single-bit weight models is only limited to simple tasks.”
BackgroundAn Energy-Efficient High Signal Margin Analog Compute-In-Memory Architecture
Kushwaha, D., Kohli, R., Mishra, J., Singh, J., Joshi, R., Dasgupta, S., & Bulusu, A. (2024) · IEEE Latin American Symposium on Circuits and Systems (LASCAS) · doi:10.1109/lascas60203.2024.10506139 · Cited by 2
“The throughput density is 32 GOPS/Kb, 64× higher than the state-of-the-art [32].”
BackgroundENOS: Energy-Aware Network Operator Search for Hybrid Digital and Compute-in-Memory DNN AcceleratorsPreprint
Nasrin, S., Shylendra, A., Kadakia, Y., Iliev, N., Gomes, W., Tulabandhula, T., & Trivedi, A. R. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2104.05217 · Cited by 2
Capacitance and Conductance Compensation Methods for Efficient Computing‐In‐Memory Designs
Luo, Y., Qiao, F., & Sun, Z. (2024) · Advanced Electronic Materials · doi:10.1002/aelm.202400452 · Cited by 1
Kim, H., An, Y., & Shim, Y. S. (2024) · International Symposium on Low Power Electronics and Design (ISLPED) · doi:10.1145/3665314.3670819 · Cited by 1
Effect of Parameter Drift on the Accuracy of Analog CNN Accelerators
Keyser, M., Ordubadi, F., & Tajalli, A. (2024) · IEEE International Midwest Symposium on Circuits and Systems (MWSCAS) · doi:10.1109/mwscas60917.2024.10658938 · Cited by 1
“The success of CNNs can be attributed to the underlying computational schemes that transform data into abstract representations, which are exploited in their deep multi-layer network architectures [2].”
BackgroundXin, G., Tan, F., Li, J., Chen, J., Yu, W.-H., Un, K.-F., Martins, R. P., & Mak, P. (2024) · IEEE International Midwest Symposium on Circuits and Systems (MWSCAS) · doi:10.1109/mwscas60917.2024.10658831 · Cited by 1
“…stores memory stationarily inside the processing engine (PE) and computes the MVM parallelly, finally reading the final or partial sum result with the corresponding circuit. which greatly reduces the costly off-chip data movement , and increases the energy efficiency and throughput [7], [8], [9].”
BackgroundHuang, S.-H., Hsiao, C.-L., & Cheng, W.-C. (2024) · Proceedings · doi:10.1109/dft63277.2024.10753500 · Cited by 1
Metadata
- Authors
- Biswas, A., & Chandrakasan, A. P.
- Affiliations
- Texas Instruments (United States) (US)
- Massachusetts Institute of Technology (US)
- Venue
- IEEE Journal of Solid-State Circuits
- Published
- 2019-01
- Volume
- 54
- Issue
- 1
- Pages
- 217-230
- ISSN
- 0018-9200 (ISSN-L)
- 1558-173X
- OpenAlex
- W2904299207
- Cited by
- 414
- References
- 29
- Citations by year
- 2026: 36 · 2025: 57 · 2024: 75 · 2023: 75 · 2022: 74 · 2021: 59 · 2020: 24 · 2019: 13
- FWCI
- 9.17
- Type
- journal article
- Language
- en
- Open access
- green
- hdl.handle.net
- Licence
- Locations
- IEEE Journal of Solid-State Circuits (publishedVersion)
- DSpace@MIT (Massachusetts Institute of Technology) (submittedVersion, cc-by-nc-sa)
- Funders
- Intel Corporation
- Topics
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Advanced Memory and Neural Computing
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Low-power high-performance VLSI design
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Analog and Mixed-Signal Circuit Design
- Keywords
- SRAM, in-memory computing, dot product computation, energy efficiency, low-power neural network hardware, analog-to-digital converter, MNIST classification
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.