Home Office NGSolve Sessions

Hi all,
Due to the ongoing situation regarding Covid19 we will not be able to hold this years User Meeting in Portland and postpone it to 2021.

In an attempt to make up for this loss we decided to move the NGSolve Sessions we held at the TU Wien in the last months to the virtual space and invite others to participate in them as well. We will also upload the sessions afterwards to our YouTube channel so that everyone has access and can rewatch them.

We have had the first of these sessions today and the material is uploaded on the channel already:

Thanks to Lukas for the very nice introduction to MPI computing with NGSolve!
If you have questions feel free to ask them in the channel or here in the forum.

If you want to participate in further sessions just say so in this topic and we will include the email address you registered here with into our invitation list.

Also if you have topics of interest you want us to discuss in one of the future sessions let us know here. We will try to include them.

We would also like to get to know our community better, so it would be nice if you could include a short description of your field of research/application and how you use NGSolve.

Best wishes and stay healthy,
Christopher

Hi Christopher,

Very nice! Please add me in.

One topic that I am interested in is a discussion on how to include alternative finite element basis functions into the library. Like the spectral element basis through Gauss-Lobatto-Legende nodes for hypercubes, or the simplicial Bernstein basis. These basis, combined with the right quadrature rules, leads to optimal complexity for high-order elements taking into account sum-factorization in the matrix/vector assembly and residual updates. I was wondering if the current high-order bases in ngsolve is competitive in terms of computational complexity.

I am trying to simulate the Talor-Green vortex problem on a cubic mesh, and found out that the computational cost of applying the convection operator for my naive high-order Taylor-Hood element is more than one order of magnitude slower than the matrix-free DG solver by Martin Kronbichler's group where they use the deal.ii package.

Best,
Guosheng

Thank you Lukas for yesterdays talk!

You can count me in for future sessions

Best wishes,
Henry

I would also be interested in future sessions!


I am particularly interested in the PETSc interface, ngsxfem and possibilities to use medical imaging data with NGSolve (e.g., uCT data, in general voxelized data).
Since I had problems compiling the PETSc interface and also NGSolve with MUMPS (some blacs / scalapack problem), I would also be interested in some more technical aspects.

Thank you for sharing the NGSolve sessions!

Best,
Julius

Hi Christopher,

Thank you for organizing such sessions. Very helpful. Please add me in.

I am interested in wave propagation and its inverse problems in general. Currently I wanted to see if I can use Netgen to simulate time-harmonic wave propagation in periodic media (for example, a waveguide with periodic refractive index). It is possible to weakly formulate the problem in W ( where the scattering medium is W-periodic) with periodic boundary conditions, however the Bloch transform appears in the bilinear form.

Regards,
Shixu

If this is still going, I'd like to participate!