E-level supercomputing ocean mode automatic transplanting optimization method and system

Abstract

The invention discloses an E-level supercomputing ocean mode automatic transplanting optimization method. The method comprises the steps of determining hot spot function code data by utilizing an established code analysis tool based on characteristics of a hot spot function; extracting hot spot function characteristics from the hot spot function code data by using a convolutional neural network, and performing vectorization processing on the hot spot function characteristics to obtain a characteristic vector of a hot spot function; inputting the feature vector of the hotspot function into a bidirectional LSTM neural network, and introducing a self-attention mechanism into the bidirectional LSTM neural network for training, so as to determine an optimal bidirectional LSTM neural network model according to the output first master-slave core code data; performing iterative training by utilizing a reinforcement learning method according to the first master-slave core code data to determinean optimal reinforcement learning model; and for the source code data acquired from the client, acquiring an optimal master-slave core code scheme by utilizing the optimal bidirectional LSTM neural network model and the optimal reinforcement learning model to realize automatic code transplantation.

Description

Technical field [0001] The present invention relates to the field of Internet technology, and more specifically, to a method and system for automatic transplantation and optimization of an E-level supercomputing ocean model. Background technique [0002] With the development of marine information technology and the national strategy of "digital ocean" and "smart ocean" centered on new generation technologies such as high-speed networks, Internet of Things, cloud computing, and artificial intelligence, marine technology has entered a period of rapid development. Human beings have higher and higher requirements for information processing capabilities, and ordinary single-core or multi-core processors can no longer meet the demand for massive computing in the ocean mode. The domestic many-core platform is a high-performance computing cluster independently developed by my country. It uses domestic many-core processors as the core. Each processor contains four heterogeneous core group...

AI Technical Summary

Problems solved by technology

[0007] The present invention proposes a method and system for automatically transplanting and optimizing E-level supercomputing ocean mo

Images