Device and method for executing neural network calculation

Abstract

The invention provides a device and method for executing neural network calculation. The device comprises an on-chip interconnection module and a plurality of neural network processing modules which are in communication with an on-chip interconnection unit, and the neural network processing modules can read and write data from other neural network processing modules through the on-chip interconnection module. In multi-core multi-layer artificial neural network calculation, neural network calculation of each layer is required to be divided, then the neural network processing modules calculate,calculation result data of each layer is obtained, and multiple neural network processing units further conduct data exchange on the calculation result data of each layer.

Description

technical field [0001] The invention belongs to the field of neural network operations, and in particular relates to a device and method for performing neural network operations. Background technique [0002] Multi-layer artificial neural networks are widely used in the fields of pattern recognition, image processing, function approximation, and optimization calculations. In recent years, multi-layer artificial networks have been favored by academic circles and researchers due to their high recognition accuracy and good parallelism. industry is getting more and more attention. Artificial neural networks involve a variety of algorithms, in which multi-core multi-layer neural network processors are used to perform neural network operations, and are widely used in various artificial neural network models and various scenarios using neural networks. [0003] One known method of supporting multi-core multi-layer neural network operations is to use a general-purpose processor. T...

AI Technical Summary

Problems solved by technology

Since the GPU is a device specially used to perform graphics and image operations and scientific computing, without special support for artificial neural network operations, a large amount of front-end decoding work is still required to perform multi-layer artificial neural network operations, which brings a lot of additional overhead.

In addition, the GPU only has a small on-chip cache, and the model data (weights) of the multi-layer artificial neural network need to be repeatedly moved from off-chip, and the off-chip bandwidth has become the main performance bottleneck.

Images