Mapreduce-based multi-GPU (Graphic Processing Unit) cooperative computing method

Abstract

The invention discloses a mapreduce-based multi-GPU (Graphic Processing Unit) cooperative computing method, which belongs to the application field of computer software. Corresponding to single-layer parallel architecture of common high-performance GPU computing and MapReduce parallel computing, a programming model adopts a double-layer GPU and MapReduce parallel architecture to help a developer simplify the program model and multiplex the existing concurrent codes through a MapReduce program model with cloud computing concurrent computation by combining the structure characteristic of a GPU plus CPU (Central Processing Unit) heterogeneous system, thus reducing the programming complexity, having certain system disaster tolerance capacity and reducing the dependency of equipment. According to the computing method provided by the invention, the GPU plus MapReduce double concurrent mode can be used in a cloud computing platform or a common distributive computing system so as to realize concurrent processing of MapReduce tasks on a plurality of GPU cards.

Description

technical field [0001] The invention relates to a multi-GPU collaborative computing method based on Mapreduce, which belongs to the field of computer software applications. Background technique [0002] In recent years, driven by hardware technology, the computing power and programmability of graphics processing units (GPUs) have been rapidly developed. The highly parallelized computing features make GPUs no longer limited to daily graphics processing tasks, and begin to involve a wider range of tasks. high-performance general-purpose computing (GPGPU) field. Because the GPU has a high-performance multi-processor array and a high-bandwidth, hidden-latency video memory system, this makes the GPU more advantageous than the traditional CPU in computing applications with a large number of repetitive data set operations and intensive memory access. Moreover, the calculation of commonly used data and programs needs to be completed by the CPU. For users, if too much CPU time is oc...

AI Technical Summary

Problems solved by technology

In the face of increasingly large-scale games and programs, the GPU acceleration of a single computer will not bring much improvement, and the contradiction between the massive growth of data and the computing power of the computer will not be resolved; the same general distributed computer Although the GPU cluster is very good in terms of computing power, once a node failure or other problems occur, the performance of the entire cluster will be greatly affected: there is also a MapReduce model that requires frequent CPU calculations when performing Map and Reduce operations, and sometimes even The CPU occupancy rate is 100%, so it is very necessary to use the participation of the GPU to balance the computing power of the system

Images