GPU video memory management control method and related device

Abstract

The invention discloses a GPU video memory management control method. The method comprises the steps of obtaining video memory use condition data of all tasks from a GPU video memory according to a preset data structure; performing placeholder transcoding on the video memory use condition data to obtain a video memory use condition placeholder; performing video memory use monitoring on the video memory use condition placeholder according to a video memory allocation table to obtain an over-allocation task; and executing a suspension operation on the over-allocation task. According to the method, condition data of the video memory is obtained from the GPU video memory, then the data is transcoded into the usage condition placeholder of the video memory, and finally the condition of the video memory is monitored according to the usage condition placeholder of the video memory, so that accurate and timely monitoring and management operation on the GPU video memory is realized, and the usage efficiency of the video memory is improved. The invention further discloses a GPU video memory management control device, a server and a computer readable storage medium, which have the above beneficial effects.

Description

technical field [0001] The present application relates to the field of computer technology, and in particular to a GPU memory management control method, a GPU memory management control device, a server, and a computer-readable storage medium. Background technique [0002] With the continuous development of information technology, in order to make the use of computers more intelligent, AI deep learning technology has emerged. Using AI deep learning technology in devices can make data processing more intelligent. Further, in order to increase the speed of deep learning, GPU (Graphics Processing Unit) is used to accelerate deep learning in the field of deep learning. However, at present, the cost of the GPU card is relatively high, and the performance of the GPU has to be fully utilized. Currently, scenarios that require the use of GPU cards include model development and model training. However, the current GPU memory is relatively large, and the model development scene or m...

AI Technical Summary

Problems solved by technology

It can be seen that in the process of using the GPU to debug the model in the prior art, due to the large GPU memory, the use of the GPU memory by each task is complicated, and it is impossible to accurately and timely monitor and manage the use of the GPU memory by the tasks in the GPU, and then The monitoring of GPU memory usage is reduced, which can easily lead to GPU memory users abusing the memory, not controlling the size of the memory, and further causing waste of GPU performance resources

Images