Turbulence Visualization at the Terascale on Desktop PCs.

Area of Science:

• Fluid Dynamics
• Computational Science

Background:

• Understanding turbulence small-scale structure and scale interactions remains a challenge.
• Visualizing high-resolution turbulent flow fields exceeds desktop computing capabilities due to memory and bandwidth limitations.

Purpose of the Study:

• To develop a GPU system for feature-based turbulence visualization overcoming current computational limitations.
• To enable interactive exploration of turbulence features in highly resolved flow fields.

Main Methods:

• A GPU system utilizing a compressed flow field representation (wavelet-based compression with run-length and entropy encoding).
• GPU-based decoding embedded into brick-based volume ray-casting.
• Derivation of turbulence properties from the velocity gradient tensor.

Main Results:

• Drastic reduction in data streamed from disk to GPU memory.
• Efficient visualization of unsteady turbulence simulations with 1024^4 spatio-temporal resolution.
• Visualization of each time step in 5 seconds on a desktop computer, with scalar feature volumes rendered three times faster.

Conclusions:

• The developed GPU system effectively visualizes complex turbulence features using compressed data.
• This approach significantly enhances the feasibility of exploring high-resolution turbulence simulations on desktop systems.
• The system advances our ability to study turbulence structure and scale interactions through interactive visualization.