Artificial neural network processor integrated circuit and design method therefor

Abstract

The invention relates to the technology of integrated circuits and the field of artificial intelligence, and especially relates to an artificial neural network processor (ANNP) integrated circuit. The integrated circuit comprises a state control module, a data control module, a parameter control module, and a nerve cell operation array module. The state control module is used for detecting the working states of the data control module, the parameter control module, and the nerve cell operation array module. After the working states of all modules are detected, the state control module generates a current process state of the whole integrated circuit, and controls the data control module, the parameter control module and the nerve cell operation array module to work synchronously. The integrated circuit is simple in structure, greatly improves the operation performance, and effectively reduces the demands for data access bandwidth. The invention also discloses an integrated circuit design method for the ANNP integrated circuit.

Description

technical field [0001] The invention relates to the field of integrated circuit technology and artificial intelligence, in particular to an artificial neural network processor (ANNP) integrated circuit and a design method of the integrated circuit. Background technique [0002] In recent years, with the popularization of the Internet and the wide application of terminal equipment sensors, the world has ushered in the era of the Internet of Things. Due to the wide application of the Internet of Everything and sensors, human beings have accumulated a large amount of data. In this era, with the simulation The artificial neural network (ANN) calculation of the neural network of the human brain has been applied in big data analysis and cloud computing, and artificial intelligence has made great breakthroughs. Microsoft Xiaobing, Apple's Siri (voice assistant), Goolge's driverless car, Japan's home robot, and AlphaGo (Go artificial intelligence), which recently defeated Li Shishi,...

AI Technical Summary

Problems solved by technology

The amount of calculation is huge. Due to the lack of a dedicated neural network processor chip, currently it mainly relies on cloud computer arrays or multi-core CPU or GPU software to complete the calculation, which is not only expensive but also consumes high power. The demand for local artificial intelligence has already begun The rise of artificial intelligence in local terminals requires low cost and power consumption to complete the calculation of artificial intelligence. Currently, the pure software method of high-end multimedia SOC (system on chip) chips is not only insufficient in calculation performance, but also high in cost and power consumption, which hinders The local application of artificial intelligence in the intelligent hardware of terminal equipment

Images