2021Unpublished venueRequires access

Real-time Super High Resolution Light Field Rendering with Multi-GPU Scheduling

Xue Liu, Xinzhu Sang, Xiao Guo, Shujun Xing, Yuanhang Li, Jinhui Yuan, Binbin Yan

Open publisher page 0 citations

Abstract

On the premise of ensuring high realistic image, it is difficult to realize real-time rendering of complex scene with ultra-high resolution 3D display. In this paper, a simple, fast and universal solution is proposed, which is a real-time method of computational integrated imaging 3D display based on multi-GPU scheduling and path tracing technology. The multi-GPU technology based on split frame rendering mode can complete several times of rendering calculation at the same time, and can reasonably adjust the proportion of each GPU rendering task through dynamic load balancing technology, so as to achieve the optimal rendering efficiency. Experimental results show that compared with the rendering frame rate of single GPU, the increase multiple of dual GPU can reach more than 1.9 in ultra-high resolution scene. In addition, for complex scene with a resolution of 7680 × 4320, the rendering frame rate of the dual GPU can reach more than 15FPS, which can basically achieve smooth interaction.

About this research paper

What this paper is about

On the premise of ensuring high realistic image, it is difficult to realize real-time rendering of complex scene with ultra-high resolution 3D display. In this paper, a simple, fast and universal solution is proposed, which is a real-time method of computational integrated imaging 3D display based on multi-GPU scheduling and path tracing technology. The multi-GPU technology based on split frame rendering mode can complete several times of rendering calculation at the same time, and can reasonably adjust the proportion of each GPU rendering task through dynamic load balancing technology, so as to achieve the optimal rendering efficiency. Experimental results show that compared with the rendering frame rate of single GPU, the increase multiple of dual GPU can reach more than 1.9 in ultra-high resolution scene. In addition, for complex scene with a resolution of 7680 × 4320, the rendering frame rate of the dual GPU can reach more than 15FPS, which can basically achieve smooth interaction.

Why it matters

A significance statement is not available in the OpenAlex record.

Key contribution

A contribution statement is not available in the OpenAlex record.

Method / approach

Method details are not available in the OpenAlex metadata.

Main findings

Findings are not separately available in the OpenAlex metadata.

Limitations

Limitations are not available in the OpenAlex metadata.

Applications

Application details are not available in the OpenAlex metadata.

Available abstract

On the premise of ensuring high realistic image, it is difficult to realize real-time rendering of complex scene with ultra-high resolution 3D display. In this paper, a simple, fast and universal solution is proposed, which is a real-time method of computational integrated imaging 3D display based on multi-GPU scheduling and path tracing technology. The multi-GPU technology based on split frame rendering mode can complete several times of rendering calculation at the same time, and can reasonably adjust the proportion of each GPU rendering task through dynamic load balancing technology, so as to achieve the optimal rendering efficiency. Experimental results show that compared with the rendering frame rate of single GPU, the increase multiple of dual GPU can reach more than 1.9 in ultra-high resolution scene. In addition, for complex scene with a resolution of 7680 × 4320, the rendering frame rate of the dual GPU can reach more than 15FPS, which can basically achieve smooth interaction.

Key concepts: Rendering (computer graphics), Computer science, Tiled rendering, Real-time rendering, Frame rate, Alternate frame rendering, 3D rendering, Path tracing

Related papers

Back to paper searchBrowse research topicsOriginal source
Real-time Super High Resolution Light Field Rendering with Multi-GPU Scheduling — Research Paper | ScholarLens