Storage and calculation integrated architecture and software optimization method for self-powered system

Abstract

The invention discloses a self-energized system oriented storage and calculation integrated architecture and a software optimization method.The self-energized system oriented storage and calculation integrated architecture comprises an energy collection and management module, a CPU module and a storage and calculation integrated module, and the output end of the energy collection and management module is electrically connected with the input end of the CPU module and the input end of the storage and calculation integrated module; the CPU module and the storage and calculation integrated module are bidirectionally and electrically connected with the STT-MRAM array, and the CPU module is arranged in the storage and calculation integrated module. According to the method, efficient operation of the binary neural network is ensured by the edge equipment through the accelerator module based on the STT-MRAM array, so that the problem of high data transmission overhead caused by the fact that the edge equipment wirelessly transmits a large amount of data to a server with relatively high performance for processing in the prior art is effectively avoided.

Description

technical field [0001] The invention relates to the field of computer architecture and storage technology, in particular to a storage-computing integrated architecture and a software optimization method for a self-powered system. Background technique [0002] Deep learning has shown remarkable performance in various intelligent applications such as natural language processing, computer vision, and speech recognition. Applying deep learning to embedded devices at the edge can make them smarter and solve a wider range of problems. The high memory capacity and computing power requirements of deep learning algorithms make complex neural network algorithms unsuitable for deployment to devices with limited resources. Moreover, for devices based on the traditional Von Neumann architecture, due to the large difference between CPU processing speed and memory access speed, the "storage wall problem" and the "power consumption wall problem" caused by a large number of data migration a...

AI Technical Summary

Problems solved by technology

However, the collected environmental energy has unstable characteristics, and there are huge challenges in how to use unstable energy

[0004] When deploying neural network algorithms on edge devices, in the past, a large amount of data was wirelessly transmitted to a computer with high performance for processing, but transmitting data requires more energy than storage or computing, and there is a certain delay in transmitting data, so this This method is not suitable for power, energy and bandwidth constrained devices

In self-powered application scenarios, the local intelligent processing of edge devices faces great challenges

Images