CPU Ray Tracing of Tree-Based Adaptive Mesh Refinement Data

Area of Science:

• Scientific Visualization
• High-Performance Computing

Background:

• Adaptive mesh refinement (AMR) is widely used in simulations but visualizing its data interactively without artifacts is difficult.
• Existing methods struggle with discontinuities at AMR level boundaries, leading to visual errors.

Purpose of the Study:

• To develop an efficient solution for direct volume rendering and hybrid implicit isosurface ray tracing of tree-based AMR (TB-AMR) data.
• To address the challenges of crack-free and artifact-free visualization of AMR data.

Main Methods:

• Proposed a novel reconstruction strategy: Generalized Trilinear Interpolation (GTI) to handle AMR level boundaries.
• Utilized a general sparse octree structure for efficient data acceleration.
• Implemented direct volume rendering and hybrid implicit isosurface ray tracing.

Main Results:

• Achieved artifact-free isosurface and volume rendering of TB-AMR data.
• Demonstrated higher quality output images compared to existing methods.
• Enabled interactive rendering rates for complex AMR datasets.

Conclusions:

• The proposed GTI strategy effectively eliminates cracks and normal discontinuities at AMR level boundaries.
• The sparse octree-based approach accelerates rendering and data queries.
• This method offers a significant improvement for interactive visualization of AMR data.