Test if FD approximation of AV derivative term can improve things
This PR adds
- the parameter
av_derivative_scheme
toScalarEq
project - a 1st order FD approximation of a 1d function
Test case for this implementation was the standard Burgers problem with a sine wave.
TL;DR: When using entropy viscosity with coefficients tuned such that enough viscosity is provided, then it seems to not matter
whether one uses the current LDG approximation or the newly implemented FD approximation.
However, using ~4000 DOFs the results obtained are far from what the FV method can yield, while the method is also incredibly fast without extra tuning or optimizations.
Furthermore, there seems to be qualitative difference in the final state when the Burger test is evolved till t=0.5
.
Still need to assess which method is correct.
This PR also did some "cosmetic" changes to ScalarEq
, e.g. re-enable callbacks and implement proper time step scaling for AV method.
And we also created a folder examples
which should be used to collect parfiles that are not meant for the ref tests.