Design and implementation of simple softcore processor using multilayer neural network on FPGA for better spatial parallelism

Abstract

This project is the Central Processing Unit (CPU) based on Neural Network (NN). Features such as making its implementation is very challenging and also very expensive, because of the large amount of hardware required. At present, cost reduction and true parallelism is very possibility to use technology such as FPGA remodeling. In this project, a soft processor cores using a multilayer neural networks on FPGA will be designed and implemented. The main reasons for using Neural Networks are as follows. It provides Massive Parallelism. The objective of this project is to design and implement Simple Soft Core Processor using Multilayer Neural Network on FPGA, to increase the processing performance by applying spatial parallelism with lower complexity and to verify design on CAD tool and board level. In this project, the Quartus II 13.1 program was introduced in the design of soft-core processor. The main factor for choosing this software because it is easy to learn and has a rapid development today. On this day, Quartus II 13.1 programming language program is among the most well-known and easy to understand. Soft-core processor in the FPGA implementation is optional software that is increasingly popular in embedded computing systems. There are several important steps that must be implemented to ensure the success of this project. Overall, the project includes the initial research to final analysis. The goals of this project are to develop a soft-core processor using multilayer neural network for better spatial parallelism.