GPU topology partitioning method and device

Abstract

The invention discloses a GPU (Graphics Processing Unit) topological partitioning method and device. The GPU topological partitioning method comprises the following steps: determining interconnectionbandwidths among a plurality of GPUs according to physical topological information of the plurality of GPUs, and generating a GPU topological graph comprising the plurality of GPUs; randomly dividinga plurality of GPUs in the GPU topological graph into two partitions; calculating migration gains of all GPUs in the GPU topological graph, migrating the GPU with the highest migration gain in the partitions including more GPUs to the partitions including less GPUs, calculating the number of cross-partition connections of the current partition scheme, and removing the migrated GPUs from the GPU topological graph; and repeating the previous step until all GPUs in the GPU topological graph are removed, and selecting a partition scheme with the minimum number of cross-partition connections as a partition result. According to the method, the topological partitions of the GPUs can be optimized from the bottom layer in a targeted mode according to different connection relations between the GPUs,the transmission time consumption between the GPUs is reduced, and the calculation speed of artificial intelligence is increased.

Description

technical field [0001] The present invention relates to the field of computers, and more specifically, to a GPU topology partitioning method and device. Background technique [0002] In the fields of high-performance computing and artificial intelligence, GPUs are often used for computing acceleration. GPU is used on a large scale due to its powerful computing power and low power consumption, especially in the hot field of artificial intelligence in recent years, most of the model training is based on GPU, which can save a lot of computing time, thus Accelerate model iteration. Due to the high cost of GPUs, more and more artificial intelligence researchers hope to fully improve the resource utilization of GPUs and maximize the value of GPUs under limited GPU resources. However, most artificial intelligence developers lack the underlying knowledge of GPUs. The communication between GPUs in the existing technology is inefficient due to the lack of underlying optimization, an...

AI Technical Summary

Problems solved by technology

However, most AI R&D personnel lack the underlying knowledge of GPU. The communication between GPUs in the existing technology is inefficient due to the lack of underlying optimization, and the lack of organization of GPU partitions slows down the calculation speed of AI.

[0003] There is no effective solution to the problem of slowing down the calculation speed of artificial intelligence caused by the lack of organization of GPU partition in the existing technology

Images