A multi-core chip thermal management method based on a recurrent neural network

Abstract

The invention relates to the technical field of information control, and discloses a multi-core chip thermal management method based on a recurrent neural network. According to the invention, a multi-core chip thermal model is established by using a recurrent neural network method, and a long-term dependency problem exists when a traditional recurrent neural network establishes a thermal model fora multi-core chip considering static power consumption; The problem is avoided by adopting an echo state network method, and an accurate thermal model is established for the multi-core chip. And thethermal model is combined with an improved model predictive control method to carry out effective thermal management on the multi-core chip. According to the method, a chip temperature is read from amulti-core chip system, a state variable is calculated by using a Kalman filter, the variable is substituted into a model prediction control method based on an echo state network, and corresponding required dynamic power input distribution is calculated according to a target temperature. According to the invention, the accurate echo state network model and the advanced model prediction control method are combined, so that the optimal effect of thermal management can be achieved.

Description

technical field [0001] The invention belongs to the field of electronic design automation, relates to the technical field of information control, and particularly relates to a multi-core chip thermal management method based on a cyclic neural network. Background technique [0002] Now, with the continuous improvement of the technology level, the power density of the device continues to increase, resulting in many heat-related problems in high-performance multi-core processors, such as system reliability problems and performance degradation. In order to find a cost-effective way to solve this problem, the researchers proposed to use dynamic thermal management to manage the thermal performance of multi-core chips through task migration and dynamic voltage frequency adjustment. In order to better guide these thermal management behaviors, dynamic thermal management methods are combined with other advanced control methods, such as traditional model predictive control methods base...

AI Technical Summary

Problems solved by technology

The first is a simple linearization of the static power consumption, which will lead to a large model error; the second is to use a polynomial approximation model, but because it is still a nonlinear model, it can only be used for a single Core Chip System

Images