Convolutional neural network calculation method for many-core processor based on double coordinate systems

Abstract

The invention discloses a multi-core processor convolutional neural network calculation method based on a dual coordinate system, relates to the field of SW heterogeneous multi-core processors and deep learning, solves the problem that the communication pressure between calculation and communication is relatively large under the existing communication constraint, and provides the following scheme. The method comprises the following steps: S1, carrying out logic filling (LP) instead of traditional physical filling (PP) on input data; s2, segmenting the logically filled data so as to distribute the data to a slave core array (CPEs) on a super computer chip; and S3, performing general matrix multiplication (iGEMM) calculation suitable for a dual coordinate system on the distributed data by the slave core array, and outputting a result. The device has the characteristics that the communication pressure between calculation and communication under the communication constraint is reduced, and the space and time complexity is reduced, so that faster matrix multiplication becomes possible.

Description

technical field [0001] The invention relates to the field of Shenwei heterogeneous many-core processors and deep learning, in particular to a convolutional neural network calculation method for many-core processors based on a dual coordinate system. Background technique [0002] Convolutional layers in convolutional neural networks extract features from the input image. Feature maps are sensitive to the location of features in the input image. A downsampling operation on the feature map is used to reduce this sensitivity. Pooling is a common downsampling operation. [0003] Pooling makes the model more robust to changes in feature locations in the input image. Pooling works similarly to convolution, sliding a kernel window over the input image and applying a pooling function to the data in the window. Two common pooling functions are: max pooling and mean pooling. The specific operations are as follows: max pooling summarizes the most active regions in the feature, ie: s...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to provide a convolutional neural network calculation method for many-core processors based on a dual coordinate system, which solves the problem of high communication pressure between calculation and communication under existing communication constraints

Images