The theme of this year’s challenge is distributed rendering, in other words, visualization of very large datasets that require parallel rendering on a cluster. To participate in the challenge, please register your interest.

Option 1: use your own data

We encourage you to use data from your own research, if you have a sufficiently large dataset. Any dataset that is too large to be rendered on a standalone desktop/workstation will qualify for this competition. A classic example is a time-dependent simulation in which disk space required to store each timestep is comparable to or exceeds the workstation’s RAM.

Option 2: use the supplied (default) dataset

If you don’t have access to such data, you can use the 3D computational fluid dynamics (CFD) dataset kindly provided for this competition by Joshua Brinkerhoff (UBC Okanagan). This dataset comes from an OpenFOAM numerical simulation of incompressible transitional air flow over a wind turbine section. The fluid is treated incompressible due to the low Mach number. The airfoil is NACA0018, a common research airfoil used to study wind turbine aerodynamics.

To reduce the size of the dataset, we include a fairly short time range () and store five variables at every third timestep, resulting in 86 steps:

  • p is the gauge static pressure (actually, static pressure divided by density, but density is constant in incompressible flow, so it functions as static pressure);
  • U is the velocity vector;
  • vorticity is the vorticity vector (curl of the velocity);
  • Lambda2 is the second eigenvalue of the symmetric tensor , where and are the symmetric and antisymmetric components of the velocity gradient tensor;
  • Q is second invariant of the velocity gradient tensor.

Please note that while the first three variables (p, U, vorticity) are available for all 86 timesteps, the last two (Lambda2, Q) are available only for the fist 50 timesteps. Also note that around , the timestep increases.

Below, we provide a simple rendering of the velocity magnitude done with ParaView on the Cedar cluster. The entire visualization with 286 frames took about 20 minutes to render on 64 CPU cores, with most time spent on the time-dependent part towards the end of the video where we had to read each timestep from disk.

(Or click here to watch this video directly on Vimeo.)

Below is a more detailed look at the separation of the laminar boundary layer from the airfoil surface and its transition to turbulence. This more complex rendering showing the isosurface of constant air speed coloured by the Y-component of the vorticity took 17 minutes on 128 CPU cores on Cedar.

(Or click here to watch this video directly on Vimeo.)

The default dataset can be found on the Cedar, Graham, Niagara, and Béluga clusters. You will need a Compute Canada account (free to all Canadian researchers) to access the data. Since the dataset is almost 1 TB in size, we stronly advise not to copy or download it to your machine. Instead, please read it directly into your visualization program from its original location.

Submissions are due by midnight Pacific Time (3 a.m. ET) on November 30th, 2019.