This page provides visualizations of some of the data in benchmark.txt, which consists of performance benchmarks for selected example cases. To add another data point, simply navigate to the build directory and run python3 benchmark.py. This will append to benchmark.txt which is tracked by git. Commit the changes if you want your benchmark to be permanently recorded; discard otherwise.

Note

The case marked "blottner sphere" is a 3D case that is set up with the Blottner sphere initial and boundary conditions, but uses a very coarse grid and a completely insufficient number of iterations. It provides a cheap assessment of 3D kernel performance, but does not measure overall time to convergence.

Overall performance

The following figure illustrates the evolution of Hexed's performance over time, accross different systems. Different series in the scatter plots indicate different machines on which the benchmarks were run.

Kernel performance breakdown

The following figures offer a more detailed breakdown of the latest benchmark on the machine ae-artl-408091. When the components don't add up to 100%, that indicates there are some subtasks whose execution time is not measured.

inviscid NACA 0012 case

viscous flat plate case

partial Blottner sphere case