Closed alexfikl closed 1 year ago
Might be worth nothing that the right preconditioned Yukawa-Beltrami operator worked with the previous settings, it's just the left-preconditioned one that was failing. This also adds the right-preconditioned ones as a slow test, since it takes about 10 minutes to run.
This is how the two versions look (the S0
IntG and friends use a Laplace kernel and the ones without a 0
use the Yukawa kernel)
LGTM! Thanks for investigating.
koelsch
), but not very satisfied with the results. Before, it was getting a small error but not converging hand now it's doing
which is a bit more converging, but not great. What do you think? I'm inclined to blame the myriad of hypersingular kernels in there.
Fixes #205