[deal.II] Re: Assembling sparse matrix from Matrix-free vmult with constrains

Hi Michal, any chance that you post or send me a small runnable test program. By the way, there is an open PR (https://github.com/dealii/dealii/pull/9343) for the computation of the diagonal in a matrix-free manner. Once this is merged, I will work the matrix-free assembly of sparse matrices.

[deal.II] Updated link to Jack Poulson's PhD thesis (in the landing page of video lecture 34)

The link to Jack Poulson's text currently in the landing page of Video Lecture 34 https://www.math.colostate.edu/~bangerth/videos.676.34.html has become out of date. For the benefit of those following the video lectures, the updated

[deal.II] Question about the advice about regularity of the stokes' equations in Step-22

Step-22 has a strong-sounding stub-like statement: * "In practice, one wants to impose as little regularity on the pressure variable as possible". * The above one-liner (fairly enough) assumes domain knowledge. As a new entrant to deal.ii (haven't studied fluid mechanics before), may I ask the

[deal.II] Re: Assembling sparse matrix from Matrix-free vmult with constrains

It cames out it is not so easy to provide simplest working example. I've tried to remake step-37, but surprisingly I was not able to reproduce the problem. I'll take a closer look what is causing a problem. W dniu wtorek, 10 marca 2020 07:46:11 UTC+1 użytkownik peterrum napisał: > > Hi Michal,

Re: [deal.II] Assembling sparse matrix from Matrix-free vmult with constrains

On 3/9/20 8:26 AM, Michał Wichrowski wrote: I've got matrix-free multigrid solver for Stokes problem. The main bottleneck is solution of coarse problem, so I tried to assemble the regular sparse matrix and use direct solver. Since the coarse problem is (relatively) small, I used vmults by unit

[deal.II] Re: solving stabilized Stokes

Hi everyone, I am back with good & bad news! - Good news first, I implemented a parallel direct solver (mumps via petsc) and going from a 2x2 blocked system to a regular one (actually 1x1 still blocked system) one just needs to adapt the setup phase, not reordering per component & not use the