MAC Unit Design For AI Accelerator

Overall, the design has been successfully verified at the simulation level and demonstrates reliable performance. The MAC unit array shows parallel processing capability, which is very important in applications such as digital signal processing, image processing, matrix multiplication, and AI hardware accelerators. In future work, the design can be implemented on FPGA hardware and further optimized to improve speed, reduce hardware area, and minimize power consumption. This would make the MAC unit array more efficient and suitable for real-time AI accelerator applications.

[1]     Umadevi, S., Penumaka, P., Ram, C. K. & Devi, T. K. High performance MAC unit design with grouping and decompositionmultiplier and 18 T gate diffusion Input-Transmission gate adder. Circuits Syst. Signal. Process. 44 (4), 2830–2854 (2025). [2]     Niknia, F. et al. A Configurable Floating-Point Fused Multiply-Add Design with Mixed Precision for AI Accelerators. IEEETransactions on Circuits and Systems for Artificial Intelligence. (2025). [3]     Dabbagh, A., Karamimanesh, M., Hassanli, K. & Abiri, E. CD-MAC: mixed signal binary/ternary in-memory computingaccelerator for power-constrained MAC processing. J. Supercomputing. 81 (1), 190 (2025). [4]     R. Sun, Y. Ni, X. He, J. Zhao, and A. Zou, ONE-SA: Enabling nonlinear operations in systolic arrays for efficient and flexibleneural network inference, in Proc. 2024 Design, Automation & Test in Europe Conf. & Exhibition, Valencia, Spain, 2024, pp. 1–6. [5]     K. Anderson, Hardware for machine learning inference: CPUs, GPUs, TPUs, https://telnyx.com/resources/ hardware-for-machine-learning-inference-cpus gpus-tpus, 2023. [6]     M.Jiang,J.Wang,A.Eldebiky,X.Yin,C.Zhuo,I.C.Lin,andG.L.Zhang,“aware pruning for efficient neural networks, ”in Design,Automation and Testin Europe(DATE),2024. [7]     J.Wang, B.Li, and G.L.Zhang, “Early exit with class exclusion for efficient in reference of neural networks,” in InternationalConference on Artificial Intelligence Circuits and Systems (AICAS),2024. [8]     Kashif Inayat, Fahad Bin Muslim, Javed Iqbal, Syed Agha Hassnain Mohsan, Hend Khalid Alkahtani and Samih M. Mostafa:“Power-Intent Systolic Array Using Modified Parallel Multiplier for Machine Learning Acceleration. Sensors”2023. [9]     Inayat, K.; Chung, J. “Hybrid Accumulator Factored Systolic Array for Machine Learning Acceleration”. IEEE Trans. VeryLarge-Scale Integration. (VLSI) Syst. 2022. [10]   Jeyakumar Ponraj, R. Jeya bharath, P. Veena, Tharumar Srihari, “High performance multiply accumulate unit by integratingbinary carry select adder and counter-based modular wall ace tree multiplier for embedding system”, the VLSI Journal 23(2023).