max_abs_speed_naive for curvilinear Euler, MHD, etc.

[andrewwinters5000]

The topic of including orientation and/or normal information for the wave speed estimates came up in this discussion.

TODO:

• compressible Euler (see fix max_abs_speed_naive for compressible Euler #888)
  • 1D (looking into the code 1D already uses the direction)
  • 2D
  • 3D
• compressible Euler multicomponent (see Fix in max_abs_speed_naive for multicomponent Euler #884)
  • 1D
  • 2D
• compressible MHD (see Adjust max_abs_speed_naive for ideal GLM-MHD #877)
  • 1D
  • 2D
  • 3D
• compressible MHD multicomponent (see Fix in max_abs_speed_naive for multicomponent MHD #883)
  • 1D
  • 2D
• acoustic perturbation 2D (see Fix in max_abs_speed_naive för APE #882)

[ranocha]

@trixi-framework/principal-developers What's you opinion on this?

[jlchan]

I'm in favor of this, and it looks like it's been done for APE already.

If incorporating orientation/normals involves the normal component of velocity, it might be possible to meet minimize how invasive the changes are by dispatching to a 1D wave speed estimate.

[ranocha]

Since I was the one asking in the thread linked above, I'm of course also in favor of this.

[andrewwinters5000]

I am also in favor of doing this. In some sense it will make the wave speed estimates still naive, but less so, because we at least account for the geometry of the element.

[ranocha]

Consensus in the meeting: Update everything; @andrewwinters5000 starts with shallow water in his PR. Then, check the list at the top.

[andrewwinters5000]

I think that the scaling used in this function for acoustic_perturbation_2d.jl

λ_max = (max(abs(v_ll), abs(v_rr)) + max(c_mean_ll, c_mean_rr)) * norm(normal_direction)

is wrong (similar to the bug fix from #831). I believe it should be

λ_max = max(abs(v_ll), abs(v_rr)) + max(c_mean_ll, c_mean_rr) * norm(normal_direction)

because the normal_direction already is scaled by the norm.

This should be fixed in this issue too.

[sloede]

Thanks for checking this out! In case you fix it also for the APE, please ping @lchristm for a review as well, such that he is aware of the change.

[andrewwinters5000]

Now that the compressible Euler branch passed tests and merged I think this issue can be closed