Memory-efficient single-pass GPU rendering of multifragment effects

Area of Science:

• Computer Graphics
• Scientific Visualization

Background:

• Rendering multifragment effects on Graphics Processing Units (GPUs) is challenging due to ordering difficulties and memory limitations.
• Existing methods struggle with large models and suffer from CPU-to-GPU data transmission bottlenecks.

Purpose of the Study:

• To present an efficient stream method for accurate rendering of multifragment effects on GPUs.
• To overcome limitations of existing methods for handling large models and improve rendering speed.

Main Methods:

• Decomposing models into parts for efficient ordering and rendering on the GPU.
• Transmitting model data part-by-part, rendering immediately, and compositing results.
• Utilizing a single data access pass with bounded memory requirements and packing parts for acceleration.

Main Results:

• The proposed stream method achieves accurate rendering of multifragment effects.
• The method significantly outperforms existing approaches in terms of speed.
• It demonstrates the capability to handle large models of any size effectively.

Conclusions:

• The stream method offers a viable solution for high-speed, accurate multifragment rendering on GPUs.
• This approach overcomes memory and data transmission limitations, enabling efficient processing of large-scale models.