tristanmontoya / stablespectralelements.jl Goto Github PK

This would make computing the pressure and velocities twice no longer necessary (this wasn't done before #56 was implemented). Maybe some way of storing $\rho V$ and $V$ to compute momentum fluxes a bit faster?

@JuliaRegistrator register

Requires simplified version of entropy projection (#62).

Even when the simulation does not run out of memory, the error calculation sometimes does. These are reasonably large 3D meshes, but I don't think error calcs should be the thing that limits how large a problem you can solve.

Currently we use the scheme's volume quadrature to evaluate the $L^2$ error. It would be nice if that could be done with higher-order quadrature.

I think this may become deprecated along with LoopVectorization.jl

@JuliaRegistrator register

This is already done for differentiation and forward/backward transform, but not for the interpolation/extrapolation operator.

I think Ranocha's flux makes most sense since it is KEP and PEP.

This will require the entropy projection (#62) , and also hybridized operators/correction terms.

Seems to manifest in conservation error being non-negligible (solution is probably off too, but discretization error masks this). Why we only see this for the 2D Euler equations is a bit of a mystery though, as the results for advection and 1D Euler are fine.

Edit: Removing inverse from metric calculation does not help, so I think it's somewhere in the Euler flux.

This may improve SIMD'ing as well.

This is probably faster since the size is known a priori.

Code is pretty fast, but compilation time is bad right now, especially in 3D. Possibly due to branches/type inference issues.

@JuliaRegistrator register

Requires adding additional connectivity fields on top of those in StartUpDG.MeshData

This issue is used to trigger TagBot; feel free to unsubscribe.

If you haven't already, you should update your TagBot.yml to include issue comment triggers.
Please see this post on Discourse for instructions and more details.

If you'd like for me to do this for you, comment TagBot fix on this issue.
I'll open a PR within a few hours, please be patient!

Start with LGL implementation.