The metrics, and how each is measured
Seven metrics. Each is stated, defined by its formula, and followed by the value measured.
Relative L2 difference
The difference of a field between the two codes over all cells, after 30 time steps with the case's own linear solvers. Median of the 42 tutorials.
(1)alpha 2.7e-12 · p_rgh 1.1e-11 · U 1.1e-08
Normalised L-infinity difference
The largest pointwise difference of the field over its scale , on the same runs. Median of the 42 tutorials.
(2)alpha 1.4e-10 · p_rgh 1.2e-10 · U 1.1e-08
Engineering quantity difference
The difference between the codes of a quantity an engineer reads off the run: the first impact pressure peak at sensor P1 at 4.89 M cells with both codes' pressure solvers converged to 1e-10, and the arrival of the front at gauge H1.
(3)0.002% (11,102.4 and 11,102.2 Pa) · arrival identical on every mesh
Mass imbalance
The change of the water volume since the start, at 6 s in the open-topped tank at 0.4 M cells, and before the jet reaches the top.
(4)−0.96% brae, −0.91% OpenFOAM · 1e-06 to 5e-06 before 0.6 s
Residual reduction
The final over the initial residual of a pressure solve, in OpenFOAM's normalisation. Median of 3,450 solves of the 42 tutorials at their own tolerances.
(5)5.4e-03 brae with the case's solvers · 5.4e-03 serial OpenFOAM
Grid convergence index
After Celik et al. (2008), on three meshes, with and the values on the fine and medium mesh, their refinement ratio and the apparent order.
(6)2.0% peak kinetic energy (order 1.37) · 0.48% arrival at H1, both codes
Speed-up
The ratio of the wall clock of the whole run to 6 s: OpenFOAM on 26 cores of a Xeon Platinum 8480+ over brae on 1 × H100.
(7)5.1× at 0.4 M · 7.3× at 1.37 M · 11.9× at 4.89 M cells