search

piernik-dev / piernik

Piernik MHD Code
GNU General Public License v3.0
15 stars 15 forks source link

Less communication in cr diffusion #436

Closed gawrysz closed 2 years ago

gawrysz commented 3 years ago

An reimplementation of #422 restricted to communication-related savings.

gawrysz commented 3 years ago

As explained in the comments to cr_diff3 the changes introduced to the solution of aMR setups aren't likely degrading the accuracy. This can be seen in crtest with problem.par modified to include some f/c boundaries. The measured error norm w.r.t. analytical solution changes by less than 1% of the error norm observed wit explicit diffusion (on current master).

gawrysz commented 3 years ago

Similar as above: the error norm in crtest changes only slightly.

gawrysz commented 2 years ago

More sophisticated approach is reqired.