GPU resource dynamic allocation method under multi-task concurrence condition

Abstract

The invention provides a GPU resource dynamic allocation method under a multi-task condition, and aims to solve the problems of lots of idle resources, reduced system throughput and unreasonable resource allocation caused by adopting a static resource allocation method during NVIDIA GPU multi-task concurrence. The method has the following three obvious characteristics: (1) configurability: a native GPU environment cannot autonomously configure the amount of occupied resources when a program runs, so that the system provides a software method, and configurability of the amount of used resources when a GPU program runs is achieved under the condition that any hardware and drive details are not modified; (2) effectiveness: the method considers affinity of tasks to different types of resources, concurrently execute the tasks with complementary resource requirements, improves the use efficiency of GPU resources, and accelerates multi-task processing; and (3) usability: the method provides a simple program conversion mode, and a developer can transfer a native program to the system to run only by adopting fixed operation steps.

Description

Technical field: [0001] The invention discloses a method for dynamically allocating GPU resources under multi-task conditions, relates to challenges faced by high-performance computing, and belongs to the technical field of computers. Background technique: [0002] As a device with a large number of computing cores that can provide high-speed parallel computing, GPU has long been used beyond the scope of graphics computing and rendering and is widely used in large-scale parallel computing such as high-performance computing, massive data processing, and artificial intelligence. As an emerging heterogeneous computing platform, many programming frameworks provide support for it, including the CUDA programming model proposed by NVIDIA, and the OpenCL programming language supported by many manufacturers. Although there is an efficient development programming model for GPU devices, the leftover resource allocation strategy adopted by GPU drivers greatly damages the parallelism of ...

AI Technical Summary

Problems solved by technology

[0009] The main purpose of the present invention is to provide a dynamic allocation method of GPU resources under the multi-task situation, to solve the problem that when the NVIDIA GPU multi-task is concurrent, a large amount of resources are idle due to the static resource allocation method, the system throughput rate decreases, and the resource allocation is unreasonable The problem has three obvious characteristics: (1) Configurability, the native GPU environment cannot independently configure the amount of resources occupied by the program when running, this system proposes a software method, without modifying any hardware and driver details (2) high efficiency, this method considers the affinity of tasks to different types of resources, and executes tasks with complementary resource requirements concurrently to improve the use efficiency of GPU resources , to speed up multitasking; (3) Ease of use, this method provides a simple program conversion mode, and developers only need to use fixed operation steps to migrate native programs to run under this system

Images