Conversation
…l vol int. dirty first implementation: fluxdifferencing volume integral is hijacked for the first implementation. at the moment only 2d
Co-authored-by: Hendrik Ranocha <ranocha@users.noreply.github.com>
ES volume integral: additions
| # TODO: This should be moved somewhere else but that would require more involved | ||
| # changes since we wouldn't know `nelements(dg, cache)`... | ||
| du_ec = Array{uEltype, 3}(undef, nvariables(equations), nnodes(dg), nelements(dg, cache)) | ||
| du_cen = similar(du_ec) | ||
| cache = (;cache..., du_ec, du_cen) |
There was a problem hiding this comment.
It would (somewhat) violate the calling convention, but could we just pass nelements(dg, cache) or cache to the corresponding create_cache method? Just a thought, I don't think this requires fixing right now, since we might switch to element-wise kernels anyways.
|
@ranocha Thanks for the update... I have to say, though, I'd very much like to find a better name for the new |
|
As a second note, I'd like to say that I am not yet sure we should implement such a flexible comparison machinery unless we have more than one application for it... Otherwise, I'd prefer to just implement a special case and then extend it to the current form once we actually do have more than one (or even more than two) different versions to compare... |
That's why I added |
Allow parameterization of VolumeIntegralLocalComparison volume integral
Co-authored-by: Michael Schlottke-Lakemper <michael@sloede.com>
…ixi.jl into hr/ho_flux_comparison
LO & HO flux comparison
Co-authored-by: Michael Schlottke-Lakemper <michael@sloede.com>
flux plus dissipation and global Lax-Friedrichs dissipation
These can be used to construct upwind SBP operators. Using the standard central DGSEM volume integral and these fluxes at the interfaces results in upwind SBP operators, which are provably linearly stable for Burgers
fluxes left and right
4 participants
TODO
weak_form_kernel!(ref, implemented in ES volume integral: additions #469)