Scalable CNN Accelarator with Error-Correcting Accumulator for High-Performance Deep Learning Applications

The Multiply and Accumulate (MAC) unit is a fundamental component in Convolutional Neural Network (CNN) accelerators, designed to enhance the efficiency and speed of deep learning applications such as image recognition, natural language processing, and autonomous driving. These applications demand high computational power and low latency, which are critical for real-time processing and decision-making. Existing solutions often struggle with power consumption, area efficiency, and processing speed, hindering their performance in resource-constrained environments. The proposed CNN-MAC accelerator addresses these challenges by integrating a Modified Booth Multiplier (MBM) and an Error Correctable Carry Look Ahead Adder (EC-CLA). The MBM enhances multiplication efficiency by reducing the number of partial products, while the EC-CLA minimizes propagation delay and corrects errors during addition, ensuring high-speed and reliable computation. This combination significantly improves the overall performance and energy efficiency of CNN accelerators, making them suitable for deployment in various high-demand scenarios.