CN119273822A - A distributed large-scale three-dimensional scene parallel rendering implementation method - Google Patents

Abstract

The invention discloses a realization method for parallel rendering of a large-scale three-dimensional scene based on a distributed type, which comprises the steps of dividing the size of a rendering target viewport according to the number of all nodes, adding a network synchronization method to synchronize rendering frames and the like. Compared with the prior art, the method has the advantages that 1) the rendering performance is remarkably improved, rendering tasks are distributed to a plurality of nodes through distributed multi-node parallel rendering, the rendering speed is greatly improved, 2) the rendering scale is expanded, the method is not limited by single machine hardware, larger-scale three-dimensional scenes can be processed, more complex and huge visual effects are displayed, 3) the utilization rate of computing resources is improved, the computing resources of the multi-nodes are fully utilized, the rendering time is shortened, and the overall efficiency is improved.

Description

Implementation method for parallel rendering of large-scale three-dimensional scene based on distribution

Technical Field

The invention relates to the technical field of three-dimensional graphics processing and rendering, in particular to a realization method for parallel rendering of a large-scale three-dimensional scene based on distribution.

Background

In the field of three-dimensional graphics rendering, along with the development of industries such as virtual reality, game development, film and television special effects and the like, higher requirements are provided for the accuracy, instantaneity and scale of three-dimensional scene rendering. Traditional large-scale three-dimensional scene rendering methods mainly rely on LOD (LevelofDetail) scene subdivision technology, and scene detail levels are dynamically adjusted to adapt to different viewing angles and performance requirements. However, the method has the obvious limitations that firstly, the number of three-dimensional models loaded on the same screen is limited, complex and huge scenes are difficult to display, and secondly, the number of rendering grids is limited by the performance of single machine hardware, so that breakthrough in effect and number is difficult to realize. In addition, single-machine rendering also faces the problems of long rendering time, low utilization rate of computing resources, and the like.

Disclosure of Invention

The invention aims to solve the technical problems that 1) the rendering performance is insufficient, the traditional method is limited by the performance of single hardware, and cannot efficiently process the rendering task of a large-scale three-dimensional scene;

2) The rendering scale is limited, namely the number of three-dimensional models loaded on the same screen is limited, and the requirements of complex scenes are difficult to meet;

3) The utilization rate of the computing resources is low, namely, the computing resources are not fully utilized when the single machine is used for rendering, so that the rendering time is long

In order to solve the technical problems, the technical scheme provided by the invention is that the realization method for parallel rendering of the large-scale three-dimensional scene based on the distribution comprises the following steps:

a1, dividing the size of the rendering target viewport according to the number of all nodes, wherein the dividing method can divide according to the same picture size, and can divide according to the rendering capability of the nodes;

A2, after the rendering flow of the three-dimensional scene is finished, adding a network synchronization method to synchronize rendering frames, and rendering the next picture after receiving an instruction, so that the consistency of all rendering result pictures is ensured;

B1, linking different rendering nodes in an Ethernet mode, and synchronizing rendering data by a multi-node computing resource R0 in a network mode;

b2, deploying the same three-dimensional scene resources on the R0, and initializing and loading the three-dimensional scene resources to confirm the requirement of network synchronization of all nodes;

the method of the synchronous network is carried out in a master-slave mode, a host is used as a synchronous service end, a slave is used as a synchronous client, and the client initializes a three-dimensional scene through the method of A2 to inform the host that data information is ready;

C2, after waiting for all the slaves to be ready by the method of B1, the host transmits a rendering instruction, the rendering instruction is transmitted to the slaves, the slaves receive the rendering instruction and then perform single-frame rendering, and the rendering frame numbers are fed back to the host through a network;

c3, the rendering frame sequences of all the slaves need to be kept consistent;

D1, the slave machine obtains a group of rendering results R0 through a C1-C3 method, wherein R0 defines a picture of a fixed viewport for each node through A1, and the picture is pushed to the host machine through a network;

And D2, the host acquires a picture set R0 through a method of D1, and acquires a complete rendering picture after filling corresponding to the picture view ports divided by A1.

Compared with the prior art, the invention has the advantages that 1) the rendering performance is obviously improved, the rendering task is distributed to a plurality of nodes through the distributed multi-node parallel rendering, and the rendering speed is greatly improved;

2) The rendering scale is expanded, the method is not limited by single machine hardware, can process larger-scale three-dimensional scenes, and shows more complex and macroscopic visual effects;

3) The utilization rate of the computing resources is improved, the computing resources of multiple nodes are fully utilized, the rendering time is shortened, and the overall efficiency is improved.

Drawings

FIG. 1 is a schematic diagram of the steps of the present invention.

Detailed Description

The present invention will be described in further detail with reference to the accompanying drawings.

In the implementation of the present invention, as shown in fig. 1, the apparatus related to the present invention mainly comprises a plurality of computing nodes, a network communication device, and a synchronization control unit. Each computing node is responsible for the rendering work of the distributed three-dimensional scene segments, the network equipment is responsible for data transmission and communication among the nodes, and the synchronous control unit is responsible for the sending and receiving of rendering instructions and the synchronization of rendering progress. The computing node is used for deploying the same three-dimensional scene resources, receiving a rendering instruction through a network, independently performing single-frame rendering, and feeding back a rendering result to the host. Network devices (e.g., ethernet switches) are used to connect the computing nodes to achieve high-speed data transmission and communication. The synchronous control unit (usually located at a host node) is used as a server and is responsible for initializing a three-dimensional scene, sending a rendering instruction, receiving rendering progress feedback, and integrating rendering results to generate a complete picture. The technical characteristics of the scheme are that a distributed rendering node mode is adopted, and the method is different from a traditional mode of arranging a plurality of computers, but a plurality of rendering nodes are deployed on one computer, and the failure or downtime of a certain node does not influence the result.

Based on the hardware structure, a matched implementation method for parallel rendering of the large-scale three-dimensional scene is provided:

a1, dividing the size of the rendering target viewport according to the number of all nodes, wherein the dividing method can divide the size of the same picture, or divide the size of the rendering target viewport according to the rendering capacity of the nodes as required, such as 1000x1000 rendering pictures, and dividing 2 rendering nodes into 500x1000 modes;

A2, after the rendering flow of the three-dimensional scene is finished, adding a network synchronization method to synchronize rendering frames, and rendering the next picture after receiving an instruction, so that the consistency of all rendering result pictures is ensured;

B1, linking different rendering nodes in an Ethernet mode, and synchronizing rendering data by a multi-node computing resource R0 in a network mode;

b2, deploying the same three-dimensional scene resources on the R0, and initializing and loading the three-dimensional scene resources to confirm the requirement of network synchronization of all nodes;

the method of the synchronous network is carried out in a master-slave mode, a host is used as a synchronous service end, a slave is used as a synchronous client, and the client initializes a three-dimensional scene through the method of A2 to inform the host that data information is ready;

C2, after waiting for all the slaves to be ready by the method of B1, the host transmits a rendering instruction, the rendering instruction is transmitted to the slaves, the slaves receive the rendering instruction and then perform single-frame rendering, and the rendering frame numbers are fed back to the host through a network;

C3, the rendering frame sequences of all the slaves need to be kept consistent, and synchronization of all nodes is carried out in a frame sequence mode of 0,1, 2;

D1, the slave machine obtains a group of rendering results R0 through a C1-C3 method, wherein R0 defines a picture of a fixed viewport for each node through A1, and the picture is pushed to the host machine through a network;

And D2, the host acquires a picture set R0 through a method of D1, and acquires a complete rendering picture after filling corresponding to the picture view ports divided by A1.

Based on the method, the whole working principle of the scheme comprises the following steps:

1) And dividing the size of the rendering target viewport according to the number of all nodes, and ensuring that each node is responsible for rendering a viewport area which is fixed or distributed according to the need.

2) And initializing and synchronizing, namely loading three-dimensional scene resources by all nodes through network synchronization, and ensuring data consistency.

3) And the rendering instruction is issued, namely the host computer is used as a synchronous server and issues the rendering instruction after waiting for all the slaves to be ready.

4) And parallel rendering, namely after receiving the rendering instruction, the slaves respectively and independently perform single-frame rendering, and feed back the rendering progress and the result to the host.

5) And integrating results, namely collecting rendering results of all the slaves by the master, and filling pictures according to the viewport division to generate a complete rendering picture.

The specific implementation mode of the method comprises the following steps:

1) And (3) environment configuration, namely installing necessary rendering software and drivers on a plurality of physical nodes, connecting all the nodes through Ethernet, and configuring network parameters.

2) And (3) scene loading, namely initializing and loading the three-dimensional scene resources on all the nodes through a synchronous control unit, and finishing loading through network synchronous confirmation.

3) And dividing the view ports, namely dividing the rendering target view ports according to the number of the nodes and the rendering capacity, and distributing the divided view ports to each node.

4) And the rendering instruction is issued, wherein the host computer sends a rendering start instruction, and each slave computer performs parallel rendering according to the allocated viewing port.

5) And integrating rendering results, namely sending the rendered frame pictures to the host computer by the slave computer through a network, and splicing the pictures by the host computer according to the view port division to generate a final rendering result.

While there has been shown and described what is at present considered to be the fundamental principles and the main features of the invention and the advantages of the invention, it will be understood by those skilled in the art that the invention is not limited to the foregoing embodiments, but is described in the foregoing description merely illustrates the principles of the invention, and various changes and modifications may be made therein without departing from the spirit and scope of the invention as hereinafter claimed. The scope of the invention is defined by the appended claims and equivalents thereof.

Claims (1)

1. A realization method for parallel rendering of a large-scale three-dimensional scene based on a distributed type is characterized by comprising the following steps:

a1, dividing the size of the rendering target viewport according to the number of all nodes, wherein the dividing method can divide according to the same picture size, and can divide according to the rendering capability of the nodes;

A2, after the rendering flow of the three-dimensional scene is finished, adding a network synchronization method to synchronize rendering frames, and rendering the next picture after receiving an instruction, so that the consistency of all rendering result pictures is ensured;

B1, linking different rendering nodes in an Ethernet mode, and synchronizing rendering data by a multi-node computing resource R0 in a network mode;

b2, deploying the same three-dimensional scene resources on the R0, and initializing and loading the three-dimensional scene resources to confirm the requirement of network synchronization of all nodes;

the method of the synchronous network is carried out in a master-slave mode, a host is used as a synchronous service end, a slave is used as a synchronous client, and the client initializes a three-dimensional scene through the method of A2 to inform the host that data information is ready;

C2, after waiting for all the slaves to be ready by the method of B1, the host transmits a rendering instruction, the rendering instruction is transmitted to the slaves, the slaves receive the rendering instruction and then perform single-frame rendering, and the rendering frame numbers are fed back to the host through a network;

c3, the rendering frame sequences of all the slaves need to be kept consistent;

D1, the slave machine obtains a group of rendering results R0 through a C1-C3 method, wherein R0 defines a picture of a fixed viewport for each node through A1, and the picture is pushed to the host machine through a network;

And D2, the host acquires a picture set R0 through a method of D1, and acquires a complete rendering picture after filling corresponding to the picture view ports divided by A1.