As seen in #194, doing grid refinement changes the mapping from reference elements to physical space, leading to a change of scale in co- and contravariant quantities like contravariant momentum h_vcon1. Consequently the convergence rate changes by one or minus one for every tensor index. Indeed hardcoding calc_error_norms(...) to use contravariant2global yields the more sensible convergence table
####################################################################################################
l2
h h_vcon1 h_vcon2
error EOC error EOC error EOC
6.17e+00 - 4.57e+03 - 4.26e+03 -
3.95e-01 3.97 6.40e+02 2.84 6.30e+02 2.76
3.36e-02 3.55 8.33e+01 2.94 8.42e+01 2.90
4.12e-03 3.03 1.08e+01 2.95 1.12e+01 2.91
mean 3.52 mean 2.91 mean 2.86
----------------------------------------------------------------------------------------------------
linf
h h_vcon1 h_vcon2
error EOC error EOC error EOC
9.17e+01 - 4.42e+04 - 3.95e+04 -
4.33e+00 4.41 5.90e+03 2.91 5.75e+03 2.78
3.38e-01 3.68 1.14e+03 2.37 1.05e+03 2.46
4.38e-02 2.95 1.66e+02 2.78 1.35e+02 2.96
mean 3.68 mean 2.68 mean 2.73
----------------------------------------------------------------------------------------------------
Note that h_vcon1 and h_vcon2 should actually read h_v1 and h_v2. Additionly I had to leave the last component h_v3 out. We should consider having global variables be the the default here.
As seen in #194, doing grid refinement changes the mapping from reference elements to physical space, leading to a change of scale in co- and contravariant quantities like contravariant momentum
h_vcon1. Consequently the convergence rate changes by one or minus one for every tensor index. Indeed hardcodingcalc_error_norms(...)to usecontravariant2globalyields the more sensible convergence tableNote that
h_vcon1andh_vcon2should actually readh_v1andh_v2. Additionly I had to leave the last component h_v3 out. We should consider having global variables be the the default here.