21 January 2022 in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
The rapid development of artificial intelligence (AI) and Internet of Things (IoT) increase the requirement for edge computing with low power and relatively high processing speed devices. The computing-in-memory (CIM) schemes based on emerging resistive nonvolatile memory (NVM) show great potential in reducing the power consumption for AI computing. However, the inconsistency of the NVM may significantly degenerate the performance of the neural network. In this article, we propose a low power resistive RAM (RRAM)-based CIM core to not only achieve high computing efficiency but also greatly enhance the robustness by bit line (BL) regulator and BL weight mapping algorithm. The simulation results show that the power consumption of our proposed 8-bit CIM core is only 12.6 mW ( $256\times 256$ at 8b). The spurious-free dynamic range (SFDR) and signal to noise and distortion ratio (SNDR) of the CIM core achieve 62.64 and 45.92 dB, respectively. The proposed BL weight mapping scheme improves the top-1 accuracy by 2.46% and 3.47% for AlexNet and VGG16 on ImageNet Large Scale Visual Recognition Competition 2012 (ILSVRC 2012) in 8-bit mode, respectively.
1 November 2023 in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
26 February 2025 in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
Computing-in-memory (CIM) has emerged as a practical paradigm to bypass the von Neumann bottleneck. However, traditional CIM schemes face challenges due to the nonideal characteristics of nonvolatile memory (NVM). To address this issue, this work provides a resistive random access memory (RRAM)-based CIM macro employing two-transistor–one-RRAM–one-capacitor (2T1R1C) cells, with capacitors reused for the successive-approximation analog-to-digital converter (SAR ADC). Single-level RRAM is utilized to mitigate resistance variation. The multiply–accumulate (MAC) operation is performed via the charge and discharge of capacitors, enhancing robustness across different process, voltage, and temperature (PVT) corners. The capacitors in 2T1R1C cells are repurposed as sampling capacitors to integrate the ADC with the array. A precision-adjustable SAR (PA-SAR) logic is proposed to generate partial sums at varying precision levels aligned with different input bits, optimizing energy efficiency while maintaining reliability. Our proposed 2T1R1C array features an average area of 3.403 μm2 for each cell, which accounts for 87.46% of the total macro area. The total macro area is 1.020 mm2 with a capacity of 256 Kb, achieving an energy density of 0.201 TOPS/mm2. The PA-SAR logic boosts energy efficiency to 44.71 TOPS/W, marking a 38.55% improvement over conventional full-precision schemes.
3 November 2023 in IEEE Transactions on Circuits and Systems II: Express Briefs
Computing-in-memory (CIM) accelerators have the characteristics of storage and computing integration, which is becoming an important solution for optimizing the computational performance of large-scale neural networks. The readout system plays a crucial role in the CIM accelerators, with power consumption accounting for a significant proportion and speed determining the overall bandwidth. In this paper, a low-power charge-domain bit-scalable readout system design is proposed for fully-parallel CIM accelerator. The charge-domain accumulator supports bit-scalable calculation modes of 2-bit, 4-bit and 8-bit for weight data by capacitor array configuration, which avoids using digital accumulators with large bitwdith. Additionally, the use of charge-domain post-weighting method for input data helps reduce power consumption with less analog-to-digital conversion. Experimental results demonstrate that the integral non-linearity (INL) and differential non-linearity (DNL) of the accumulator output voltage VOUT of in each mode do not exceed 1 LSB, ensuring excellent linearity and accurate bit-scalable MAC calculations. Moreover, the proposed charge-domain CIM accelerator with ReRAM macro cells achieves an average energy efficiency of 180.61 TOPS/W and a normalized peak energy efficiency of 1618.0 TOPS/W on mixed-precision ResNet-18 networks.
13 December 2021 in IEEE Transactions on Circuits and Systems II: Express Briefs
The ever-increasing computation complexity of fast-growing Deep Neural Networks (DNNs) has requested new computing paradigms to overcome the memory wall in conventional Von Neumann computing architectures. The emerging Computing-In-Memory (CIM) architecture has been a promising candidate to accelerate neural network computing. However, data movement between CIM arrays may still dominate the total power consumption in conventional designs. This brief proposes a flexible CIM processor architecture named Domino and “Computing-On-the-Move” (COM) dataflow, to enable stream computing and local data access to significantly reduce data movement energy. Meanwhile, Domino employs customized distributed instruction scheduling within Network-on-Chip (NoC) to implement inter-memory computing and attain mapping flexibility. The evaluation with prevailing DNN models shows that Domino achieves 1.77-to- $2.37\times $ power efficiency over several state-of-the-art CIM accelerators and improves the throughput by 1.28-to- $13.16\times $ .
15 September 2022 in IEEE Journal on Exploratory Solid-State Computational Devices and Circuits
The growing data volume and complexity of Deep Neural Networks (DNN) require new architectures to surpass the limitation of the von-Neumann bottleneck, with Computing-in-memory (CIM) as a promising direction for implementing energy-efficient neural networks. However, CIM’s peripheral sensing circuits are usually power and area hungry components. We propose a time-multiplexing Computing-in-Memory architecture (TM-CIM) based on memristive analog computing to share the peripheral circuits and process one column at a time. The memristor array is arranged in a column-wise manner that avoids wasting power/energy on unselected columns. In addition, DAC (digital-to-analog converter) power and energy efficiency, which turns out to be an even greater overhead than ADC (analog-to-digital converter), can be fine-tuned in TM-CIM for significant improvement. For a 256*256 crossbar array with a typical setting, TM-CIM saves 18.4× in energy with 0.136 pJ/MAC efficiency, and 19.9× area for 1T1R case and 15.9× for 2T2R case. Performance estimation on VGG-16 indicates that TM-CIM can save over 16× area. A trade-off between the chip area, peak power, and latency is also presented, with a proposed scheme to further reduce the latency on VGG-16, without significantly increasing chip area and peak power.