Terminal equipment heterogeneous processor inference acceleration method under temperature constraint

Abstract

The invention provides a terminal equipment heterogeneous processor inference acceleration method under temperature constraint, aims at intelligent terminal equipment equipped with a plurality of heterogeneous processors in an industrial production environment, and solves the problem of low terminal equipment inference efficiency caused by deep neural network interlayer heterogeneity, processor heterogeneity and environment temperature. According to the method, firstly, the environment temperature of industrial production and the power of a terminal device processor are considered, a terminal device dynamic frequency model under temperature constraint is established, and a temperature sensing dynamic frequency algorithm is used for setting the device frequency; secondly, designing a deep neural network single-layer parallel method according to calculation modes and structural characteristics of different layers in the deep neural network; and finally, a heterogeneous processor in the terminal equipment is utilized to design a deep neural network single-layer calculation task allocation method oriented to the heterogeneous processor, and low delay and robustness of collaborative inference of the heterogeneous processor of the terminal equipment are guaranteed.

Description

Technical field [0001] The present invention belongs to the field calculation and industrial Internet, which specifically involves a terminal apparatus heterogeneous processor inferred acceleration method under temperature constraints. Background technique [0002] With the rapid development of hardware technology, there are currently, smart phones, smart cameras, wearable devices, drones, etc., have been widely used in various fields of people's production and life, and these intelligent equipment are often equipped with a camera. Radar, etc. Neural network processor, which accelerates artificial intelligence algorithm, and the like. These processors have strong computing power, and after the sensor acquires information, sounds, etc., the acquired data can be analyzed. On the other hand, the number of depth neural networks is constantly deepening, the amount of data and calculation is increasing; on the other hand, most of the intelligent terminal equipment is limited by power c...

AI Technical Summary

Problems solved by technology

However, on the one hand, existing smart terminal devices are often difficult to meet the inference accuracy and delay requirements of deep neural network applications due to their poor overall computing performance; on the other hand, the complexity of the industrial production environment affects the performance of smart terminal devices. There are also many restrictions

Therefore, high accuracy and inference delay requirements and relatively weak terminal equipment computing power limited by the industrial production environment pose new challenges to the traditional deep neural network inference method based on intelligent terminal equipment.

Images