MZMufeng Zhou

Mufeng Zhou

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Mufeng Zhou has a total of 38 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • DCiROM: A High-Density Fully-Digital Compute-in-Read-Only-Memory Macro for Energy-Efficient Task-Level DNN Inference

    by Tianyi Yu, Tianyu Liao, Mufeng Zhou, Xiaotian Chu, Guodong Yin, Mingyen Lee, Yongpan Liu, Huazhong Yang, Xueqing Li

    Compute-in-memory (CiM) is a promising approach to alleviating the von Neumann memory wall bottleneck in data-intensive tasks. As a candidate to build CiM, static-random-access memory (SRAM) has become popular in accelerating deep neural networks (DNNs) because of its mature fabrication support and high flexibility. However, SRAM-based CiM suffers from low memory density, resulting in frequent weight reloading and increased power consumption during end-to-end inference tasks. Recently, read-only-memory (ROM)-based analog CiM (ACiM) has demonstrated the potential to deploy all DNN weights on-chip by leveraging high-density ROM. It may also achieve encouraging flexibility by exploiting the synergy between a small SRAM and a big ROM. Nevertheless, it faces challenges in achieving high computing density due to limitations in analog-to-digital converter (ADC) performance. Meanwhile, digital CiM (DCiM) improves computing density but incurs significant energy and area overhead due to the adder tree. To address these challenges, this paper proposes the first fully digital and flexible compute-in-read-only-memory (DCiROM) design approach. On the one hand, DCiROM introduces a novel ROM-logic fusion CiM that reduces the CiM area by 51% while maintaining high memory density and computing performance. On the other hand, DCiROM achieves high flexibility with minimal area overhead by proposing two innovative multiply-and-accumulate (MAC) resource-reusing techniques. This work has implemented a DCiROM chip with 3024Kb ResNet-56 parameters using a 65nm CMOS technology. This macro achieves 6.9-29.5x higher normalized FoM (memory density x computing density) than the state-of-the-art CiM works. Additionally, it reduces energy consumption per image inference by 2.9-9.9x compared to SRAM-based CiM works when off-chip access is considered.


  • Hybrid SRAM/ROM Compute-in-Memory Architecture for High Task-Level Energy Efficiency in Transformer Models With 8928-kb/mm² Density in 28nm CMOS

    by Guodong Yin, Yiming Chen, Mingyen Lee, Xirui Du, Yue Ke, Wenjun Tang, Zhonghao Chen, Mufeng Zhou, Jinshan Yue, Huazhong Yang, et al.

    This article reports the first ultra-high-density hybrid static random access memory (SRAM)/read-only memory (ROM) compute-in-memory (CiM) architecture for multibit multiply-accumulate (MAC) operations of transformer models. It features several techniques that enhance the on-chip weight density, energy efficiency, and flexibility, namely: 1) the hybrid CiM structure of 6T SRAM and 1T multi-level cell (MLC) ROM for ultra-high memory density to improve the limited on-chip memory capacity and alleviate the frequent weight reload from dynamic random access memory (DRAM); 2) 2-/5-b adaptive-resolution analog-to-digital converters (ADCs) for energy reduction and low accuracy loss; and 3) post-fabrication 1-of-4 capacitor selection (PFCS) to reduce the capacitance of computing capacitors, which was originally limited by the capacitor variation. This leads to lower energy and higher accuracy. A record-high 22-Mb ROM CiM and 896-kb SRAM CiM macro has demonstrated the 8 × 8 b MAC operations of one transformer layer, featuring ultra-high weight density of 8928 kb/mm2 and significantly reduced DRAM access activities toward high inference energy efficiency.


  • Open Access

    DCiROM: A Fully Digital Compute-in-ROM Design Approach to High Energy Efficiency of DNN Inference at Task Level

    by Tianyi Yu, Tianyu Liao, Mufeng Zhou, Xiaotian Chu, Guodong Yin, Mingyen Lee, Yongpan Liu, Huazhong Yang, Xueqing Li

    Owing to mature fabrication support and high flexibility, static random-access memory (SRAM) has become a very promising candidate for compute-in-memory (CiM) in accelerating deep neural networks (DNNs). However, SRAM-based CiM has low memory density and thus very limited total on-chip capacity, resulting in frequent weights reloading and additional power consumption during end-to-end inference tasks. Analog ROM CiM increases memory density but suffers from low computing density caused by A/D converter (ADC) limitation. To address these challenges, for the first time, a fully digital compute-in-read-only-memory (DCiROM) design approach is proposed in this paper. DCiROM introduces a novel ROM-logic fusion CiM that successfully reduces CiM area by 51% while maintaining high memory density and computing performance. By reusing multiply-and-accumulation (MAC) resources, DCiROM further achieves flexibility with a minimal area cost. We have implemented a DCiROM chip loaded 3024Kb ResNet-56 parameters using 65nm CMOS technology. This macro achieves 10.2x-55.7x higher normalized FoM (memory density x computing density) than the state-of-the-art CiM works. It also reduces 2.9x-9.9x energy consumption per image inference than SRAM CiM works when considering off-chip access.


  • A 28-nm Floating-Point Computing-in-Memory Processor Using Intensive-CIM Sparse-Digital Architecture

    by Shengzhe Yan, Jinshan Yue, Chaojie He, Zi Wang, Zhaori Cong, Yifan He, Mufeng Zhou, Wenyu Sun, Xueqing Li, Chunmeng Dou, et al.

    Computing-in-memory (CIM) chips have demonstrated promising high energy efficiency on multiply–accumulate (MAC) operations for artificial intelligence (AI) applications. Though integral (INT) CIM chips are emerging, the floating-point (FP) CIM chip has not been well explored. The high-accuracy demand of larger models and complex tasks requires FP computation. Besides, most of the neural network (NN) training tasks still rely on FP computation. This work presents an energy-efficient FP CIM processor. It is observed that most of the exponent values of FP data are concentrated in a small region. Therefore, the FP computations are divided into intensive and sparse parts and then executed on an intensive-CIM sparse-digital architecture. First, an FP-to-INT CIM workflow for the intensive FP operations is designed to reduce the CIM execution cycles. Second, a flexible sparse-digital core is proposed for the remaining sparse FP operations. Utilizing both the intensive-CIM and sparse-digital cores, this work can achieve both high energy efficiency and identical accuracy to the FP algorithm baseline. Considering the FP CIM execution flow, a CIM-friendly low-bit FP training method is proposed to further reduce the execution cycles. Besides, a low-MAC-value (MACV) CIM macro is designed to utilize the more random sparsity brought by FP alignment. The 28-nm fabricated chip shows 275–1615-TOPS/W@INT4 and 17.2–91.3-TOPS/W@FP16 macro energy efficiency from dense to the average sparsity on the tested models.


  • Cramming More Weight Data Onto Compute-in-Memory Macros for High Task-Level Energy Efficiency Using Custom ROM With 3984-kb/mm2 Density in 65-nm CMOS

    by Guodong Yin, Yiming Chen, Mufeng Zhou, Wenjun Tang, Mingyen Lee, Zekun Yang, Tianyu Liao, Xirui Du, Vijaykrishnan Narayanan, Huazhong Yang, et al.

    Owing to the mature process and low access energy, static random-access memory (SRAM) has become a promising candidate for compute-in-memory (CiM) acceleration of multiply-accumulate (MAC) operations. However, SRAM-based CiM cells have rather low density and thus very limited total on-chip memory capacity. This fact, unfortunately, results in undesired weight data reload operations from the off-chip dynamic random-access memory (DRAM) in data-intensive scenarios and may even tarnish the energy efficiency of CiM at the task level. Therefore, exploration toward higher density CiM in CMOS is critical to ensure truly high energy efficiency in practice. Aligned with the goal of ultrahigh density, this article presents the first one-transistor (1T) multi-level-cell (MLC) read-only memory (ROM) CiM macro for multi-bit MAC. The highlights of the proposed ROM CiM techniques include: 1) multi-source-driven (MSD) 1T-MLC ROM; 2) charge-domain capacitor sharing (CDCS) for ultrahigh CiM memory density; and 3) ROM-based transfer-learning architectures to provide flexible support of different tasks with minor accuracy degradation. These techniques are demonstrated with a fabricated 2-Mb 1T-MLC ROM CiM macro for 8 b $\times$ 8 b MAC computing. This macro features a record-high cell density of 0.096- $\mu$ m $^{2}$ /bit and a macro weight density of 3984 kb/mm $^{2}$ in a 65-nm pure CMOS technology. It also achieves 3.8 $\times$ –55.3 $\times$ lower energy consumption per image inference than the state-of-the-art CiM macros when considering the possible DRAM access.


  • GRAPHIC: Gather and Process Harmoniously in the Cache With High Parallelism and Flexibility

    by Yiming Chen, Mingyen Lee, Guohao Dai, Mufeng Zhou, Nagadastagiri Challapalle, Tianyi Wang, Yao Yu, Yongpan Liu, Yu Wang, Huazhong Yang, et al.

    In-memory computing (IMC) has been proposed to overcome the von Neumann bottleneck in data-intensive applications. However, existing IMC solutions could not achieve both high parallelism and high flexibility, which limits their application in more general scenarios: As a highly parallel IMC design, the functionality of a MAC crossbar is limited to the matrix-vector multiplication; Another IMC method of logic-in-memory (LiM) is more flexible in supporting different logic functions, but has low parallelism. To improve the LiM parallelism, we are inspired by investigating how the single-instruction, multiple-data (SIMD) instruction set in conventional CPU could potentially help to expand the number of LiM operands in one cycle. The biggest challenge is the inefficiency in handling non-continuous data in parallel due to the SIMD limitation of (i) continuous address, (ii) limited cache bandwidth, and (iii) large full-resolution parallel computing overheads. This paper presents GRAPHIC, the first reported in-memory SIMD architecture that solves the parallelism and irregular data access challenges in applying SIMD to LiM. GRAPHIC exploits content-addressable memory (CAM) and row-wise-accessible SRAM. By providing the in-situ, full-parallelism, and low-overhead operations of address search, cache read-compute-and-update, GRAPHIC accomplishes high-efficiency gather and aggregation with high parallelism, high energy efficiency, low latency, and low area overheads. Experiments in both continuous data access and irregular data pattern applications show an average speedup of 5x over iso-area AVX-like LiM, and 3-5x over the emerging CAM-based accelerators of CAPE and GaaS-X in advanced techniques.


  • SAMBA: Single-ADC Multi-Bit Accumulation Compute-in-Memory Using Nonlinearity- Compensated Fully Parallel Analog Adder Tree

    by Yiming Chen, Guodong Yin, Mufeng Zhou, Wenjun Tang, Zekun Yang, Mingyen Lee, Xirui Du, Jinshan Yue, Jiaxin Liu, Huazhong Yang, et al.

    Performing data-intensive tasks in the von Neumann architecture is challenging to achieve both high performance and energy efficiency due to the memory wall bottleneck. Compute-in-memory (CiM) is a promising mitigation approach by enabling parallel and in-situ multiply-accumulate (MAC) operations within the memory array. Thanks to the good matching of capacitors, SRAM-based charge-domain CiM (Q-CiM) has shown its potential for higher row-wise parallelism. However, the peripheral circuits of Q-CiM, such as the input drivers and analog-digital converters (ADCs), limit further improvement of throughput and area efficiency. This paper proposes a single-ADC multi-bit accumulation CiM macro architecture SAMBA, which can perform multi-bit MAC operation with ReLU of two vectors in one CiM cycle by only a single A/D conversion to mitigate the ADC overhead. In addition, post-correction methods are proposed to compensate the non-linearity of sensitive circuit modules in SAMBA to recover the accuracy drop due to the capacitor mismatch. A proof-of-concept macro is fabricated in a 65nm process and achieves 51.2GOPS throughput and 10.3TOPS/W energy efficiency, while showing 88.6% accuracy on CIFAR-10 and 64.8% accuracy on the CIFAR-100 with VGG-8 model.


  • A Heterogeneous Microprocessor Based on All-Digital Compute-in-Memory for End-to-End AIoT Inference

    by Songming Yu, Yifan He, Hongyang Jia, Wenyu Sun, Mufeng Zhou, Luchang Lei, Wentao Zhao, Guofu Ma, Huazhong Yang, Yongpan Liu

    Deploying neural network (NN) models on Internet-of-Things (IoT) devices is important to enable artificial intelligence (AI) on the edge realizing AI-of-Things (AIoT). However, high energy consumption and bandwidth requirement of NN models restricts AI applications on battery-limited equipments. Compute-In-Memory (CIM), featured with high energy efficiency, provides new opportunities for the IoT deployment of NN. However, the design of CIM-based full system is still at the early stage, lacking system-level demonstration and vertical optimization for running end-to-end AI applications. In this paper, we demonstrate a low-power heterogeneous microprocessor System-on-Chip (SoC) with an all-digital SRAM CIM accelerator and rich data acquisition interfaces for end-to-end AIoT NN inference. A dedicated reconfigurable dataflow controller for CIM computation greatly lowers bandwidth requirement on the system bus and improves execution efficiency. The all-digital SRAM CIM array embeds NAND-based bit-serial multiplication within the readout sense amplifier balancing the storage density and system-level throughput. Our chip achieves a throughput of 12.8 GOPS, with 10 TOPS/W energy efficiency. Benchmarked by the four tasks in MLPerf Tiny, experimental results show 1.8x to 2.9x inference speedup over a baseline CIM processor.