koomie
commented
2 years ago Documenting the order of the solution differences incurred:
density
$ h5diff -r --delta=1e-9 ref_solns/cyl3d_coarse.4iters.h5 ref_solns_pressure_primitives/cyl3d_coarse.4iters.h5 /solution/density
dataset: </solution/density> and </solution/density>
size: [24612] [24612]
position density density difference
------------------------------------------------------------
[ 20040 ] 1.21327 1.21327 1.01839e-09
[ 20721 ] 1.21415 1.21415 1.11728e-09
[ 22185 ] 1.21718 1.21718 1.5326e-09
[ 22425 ] 1.21772 1.21772 1.61963e-09
[ 23360 ] 1.21595 1.21595 1.27625e-09
[ 23398 ] 1.2147 1.2147 1.18119e-09
6 differences foundrho-u
$ h5diff -r --delta=1e-6 ref_solns/cyl3d_coarse.4iters.h5 ref_solns_pressure_primitives/cyl3d_coarse.4iters.h5 /solution/rho-u
dataset: </solution/rho-u> and </solution/rho-u>
size: [24612] [24612]
position rho-u rho-u difference
------------------------------------------------------------
[ 22426 ] 18.0379 18.0379 1.02457e-06
[ 22427 ] 17.1003 17.1003 1.10245e-06
2 differences foundrho-v
$ h5diff -r --delta=2e-7 ref_solns/cyl3d_coarse.4iters.h5 ref_solns_pressure_primitives/cyl3d_coarse.4iters.h5 /solution/rho-v
dataset: </solution/rho-v> and </solution/rho-v>
size: [24612] [24612]
position rho-v rho-v difference
------------------------------------------------------------
[ 5875 ] 0.837575 0.837575 2.14437e-07
[ 6867 ] -0.0616651 -0.0616653 2.13152e-07
[ 10861 ] -0.0729642 -0.0729639 2.06637e-07
[ 19091 ] 0.182275 0.182275 2.04475e-07
[ 19576 ] 1.06406 1.06406 2.05506e-07
[ 20018 ] 1.21144 1.21144 2.5861e-07
[ 20507 ] 1.06645 1.06645 2.02288e-07
[ 20522 ] -1.22843 -1.22843 3.15595e-07
[ 21356 ] -0.967231 -0.967231 2.84834e-07
[ 21702 ] -1.50662 -1.50662 3.09351e-07
[ 21725 ] 0.615428 0.615428 2.09092e-07
[ 22104 ] -0.498491 -0.498491 2.19315e-07
[ 22209 ] 0.42436 0.42436 2.6855e-07
[ 22210 ] -0.0749706 -0.0749708 2.08657e-07
[ 22278 ] 0.507382 0.507382 3.20415e-07
[ 22489 ] -0.824939 -0.82494 3.38774e-07
[ 22602 ] -0.486034 -0.486034 3.67441e-07
[ 22824 ] 0.904103 0.904103 3.87538e-07
[ 23360 ] 0.00346558 0.00346537 2.13122e-07
[ 24548 ] 0.661496 0.661496 2.10979e-07
20 differences foundrho-w
$ h5diff -r --delta=2e-7 ref_solns/cyl3d_coarse.4iters.h5 ref_solns_pressure_primitives/cyl3d_coarse.4iters.h5 /solution/rho-w
dataset: </solution/rho-w> and </solution/rho-w>
size: [24612] [24612]
position rho-w rho-w difference
------------------------------------------------------------
[ 3267 ] 0.108117 0.108117 2.33845e-07
[ 9462 ] 0.110773 0.110773 2.01423e-07
[ 17523 ] -0.292629 -0.292629 2.16004e-07
[ 18810 ] -1.08976 -1.08976 2.76116e-07
[ 20338 ] -0.734929 -0.73493 3.10586e-07
[ 20339 ] -0.549149 -0.54915 2.11404e-07
[ 20506 ] 1.78762 1.78762 2.2018e-07
[ 20722 ] -0.793232 -0.793232 2.2132e-07
[ 20823 ] 1.71891 1.71891 2.20417e-07
[ 20892 ] -0.644418 -0.644418 2.27947e-07
[ 21068 ] -0.0345053 -0.0345055 2.54667e-07
[ 21109 ] 1.13453 1.13453 2.90457e-07
[ 21141 ] -0.0135222 -0.0135219 2.45357e-07
[ 21143 ] -0.142196 -0.142196 2.4571e-07
[ 21349 ] 0.418434 0.418434 2.80406e-07
[ 21350 ] 0.0204053 0.0204051 2.37715e-07
[ 21412 ] -0.827687 -0.827687 2.35675e-07
[ 21557 ] 0.15843 0.15843 2.41911e-07
[ 21558 ] -0.234227 -0.234228 2.46029e-07
[ 21737 ] -1.01409 -1.01409 2.20639e-07
[ 21842 ] 1.39407 1.39407 3.56239e-07
[ 21990 ] 1.54663 1.54663 2.33166e-07
[ 22202 ] 0.324418 0.324418 2.05519e-07
[ 22426 ] -1.49499 -1.49499 2.59193e-07
[ 22576 ] -0.161705 -0.161705 2.12475e-07
[ 23360 ] -1.04003 -1.04003 2.26341e-07
[ 23412 ] 1.22676 1.22676 2.35979e-07
[ 23414 ] -0.887601 -0.887601 2.02283e-07
28 differences foundrho-E (relative)
$ h5diff -r --relative 4e-9 ref_solns/cyl3d_coarse.4iters.h5 ref_solns_pressure_primitives/cyl3d_coarse.4iters.h5 /solution/rho-E
dataset: </solution/rho-E> and </solution/rho-E>
size: [24612] [24612]
position rho-E rho-E difference relative
------------------------------------------------------------------------
[ 20040 ] 258408.7971 258408.7961 0.001047896454 4.05518874e-09
[ 20505 ] 258520.7132 258520.712 0.001172940305 4.537123122e-09
[ 20520 ] 257966.4086 257966.4074 0.001264759427 4.902806716e-09
[ 20821 ] 258562.9616 258562.9605 0.001126207644 4.355641847e-09
[ 21841 ] 257994.3599 257994.3586 0.001231313247 4.772636286e-09
[ 22185 ] 259677.9486 259677.9474 0.001181709435 4.550673061e-09
[ 22425 ] 259534.9274 259534.9259 0.001497964869 5.771727429e-09
[ 22427 ] 270698.4881 270698.4894 0.001301015902 4.806143927e-09
[ 23116 ] 269380.8345 269380.8358 0.001325124467 4.919149088e-09
[ 23117 ] 258287.227 258287.2258 0.00122206597 4.731422365e-09
[ 23201 ] 258496.0412 258496.0402 0.001045500365 4.044550805e-09
[ 23360 ] 259471.2188 259471.2175 0.001323732024 5.101652656e-09
[ 23362 ] 270187.2363 270187.2375 0.001228933688 4.548452059e-09
[ 23398 ] 259474.1805 259474.1795 0.001072542917 4.133524634e-09
[ 23412 ] 258643.9934 258643.9924 0.001051878935 4.066898707e-09
[ 23413 ] 269417.8694 269417.8705 0.001091087761 4.049797386e-09
16 differences found
Changes
This PR contains 3 main changes:
GasMixturehas been created that allows for more generic fluids. Its capabilities have not yet been fully decided but it acts as a base class for further development.The GPU version has also been updated. It is now able to run with passive scalar. The drawback is that now it is slower. This is because in some kernels, the interpolation was carried out after loading all required data for a particular element. This is no longer possible for increased number of equations (it is now limited to 20 equations). Interpolation is now carried out for each variable which means interpolants are loaded as many times as equations are solved. This decrease in performance can potentially offset by running with higher polynomial orders (this last comment to be checked).
Things to bear in mind
DryAirin places. This will need to be revisited once we freeze theGasMixtureclassMASA convergence tests
Convergence rates have been verified with MASA for both Euler (Re=\infty) and and Navier-Stokes with Re=14. The Mach number in both cases was M=0.1. While the rates remain the same for the Euler case (to be expected since these equations are independent of temperature gradients) the rates change slightly for the Navier-Stokes case (see below). This is also to be expected due to the current implementation of the viscous fluxes.