GPU processing method and device for data calculation flow graph

Abstract

The invention provides a GPU (Graphics Processing Unit) processing method and device for a data calculation flow graph. The method comprises the following steps: storing received input data into a first video memory block distributed for a source node in the calculation flow graph; obtaining an expression of a subsequent node pointed by the source node, reading input data from the first video memory block, and distributing a second video memory block for the subsequent node according to the video memory block distributed for the source node in the calculation flow graph; under the condition that all the data needing to be input of the expressions of the subsequent nodes are obtained, inputting the read input data into the expressions to be calculated, obtaining a result of the subsequent nodes, and storing the result of the subsequent nodes in a second video memory block; and reading a result of the subsequent node from the second video memory block, obtaining a result of processing the input data according to the calculated flow graph according to the read result of the subsequent node, and outputting a processing result. By means of the scheme, it can be avoided that data are copied back and forth between the GPU and the external device, and therefore the data processing speed of the GPU is increased.

Description

technical field [0001] The present invention relates to the technical field of data processing, in particular to a GPU (Graphics Processing Unit, graphics processor) processing method and device for a data calculation flow graph. Background technique [0002] In real life, there is a wide range of sequence data (for example, time series, which can be a series of values ​​obtained by observing a certain physical quantity at equal intervals), covering many fields such as finance, medicine, biology, and chemistry. The processing of sequence data often involves cumbersome and complicated calculation processes, and these calculation processes can have good data parallelism, that is, the same operation can be performed on many data at the same time. [0003] The processing of sequence data can be represented by a computational flow graph. Computational flow graphs can clearly express the computational process of cumbersome and complex sequence data. In the calculation flow graph...

AI Technical Summary

Problems solved by technology

In the process of calculating the results of the nodes, data needs to be copied back and forth between the GPU and the CPU, and the process of copying data is time-consuming, which will reduce the speed of highly parallel processing sequence data on the GPU.

Images