Method for balancing dynamic feedback load based on PKDT (Prediction KD Tree) in cluster rendering environment

Abstract

The invention discloses a method for balancing a dynamic feedback load based on a PKDT (Prediction KD Tree) in a cluster rendering environment. The method comprises the following steps that specific to a three-dimensional model required to be rendered, a server partitions each rendered frame of task into a plurality of subtasks, pre-assigns the subtasks to each rendering machine node, and transmits one by one; (2) after executing one subtask, each rendering machine node returns a rendering result image to a server machine node, and applies for a next subtask; and (3) the server records time when each subtask is completed by using each rendering machine node and the quantity of the subtasks completed by using each rendering machine node, and performs dynamic balanced distribution on uncompleted subtasks among rendering machine nodes according to a recording result till the frame of task is rendered. According to the method for balancing the dynamic feedback load based on the PKDT in the cluster rendering environment, each frame of rendering task is partitioned uniformly according to a screen space, and the partitioned subtasks are dynamically and evenly distributed in each rendering machine node, so that the optimal rendering capability of the cluster rendering system is excavated fully.

Description

technical field [0001] The invention relates to the field of cluster rendering, in particular to a dynamic feedback load balancing method based on a PKDT tree in a cluster rendering environment. Background technique [0002] At present, high-realism rendering has been used more and more in many fields, such as the production of special effects in movies and the display of realistic scenes. [0003] High-realistic rendering requires extremely high computing power. If there is a high requirement for rendering speed, then the rendering capability of a single machine can no longer be satisfied. Second, cluster technology is required. [0004] Load balancing is one of the important factors affecting the overall performance of the cluster rendering system. In a cluster rendering system, due to differences in processing capabilities between different rendering nodes or uneven task distribution, there will be inconsistencies in the completion time of tasks on different rendering no...

AI Technical Summary

Problems solved by technology

This method has great limitations. First of all, it is usually difficult to accurately estimate the computing power and network environment of rendering nodes. Secondly, since the computing environment and network environment of each rendering node are constantly changing, the calculation of these rendering nodes Capabilities and network transmission capabilities are constantly changing, so this method is usually difficult to achieve good results

[0007] The dynamic balance method is currently a popular method. There are many algorithms of this type, such as the Diffusion method, the Median-cut algorithm, the Master-slave method, etc., but there is still no mature dynamic load balancing method that can be used in a cluster rendering environment by screen way of dividing space

Images