The flux limiter scheme has caused us lots of inexplicable slowdowns, and we were waiting with the merge until we resolved that along with the X-point temperature spike bug. With those two problems having turned out to be challenging, I think we should fix the equations in a separate PR as they are simply not correct. We can then tackle the limiter scheme methodically bit by bit.
I had to merge the other 2D neutral fixes PR into this as both are required for the tests. That other PR is causing checkerboarding in the neutral_mixed integrated test.
To do:
[x] Implement the equation fixes: kappa and eta factors, perpendicular advection and conduction factors
[x] Implement the viscous heating
[x] Test and record impact on results and performance
[ ] Update docs
Original work from @bendudson:
Impact on results is not insubstantial, especially at the target. Performance is the same or slightly faster - probably dependent on the change in conditions rather than being more difficult numerically.
There is an outstanding PR which has fixes for neutral equations as well as the new limiter scheme. It lives on my fork at https://github.com/mikekryjak/hermes-3/pull/5.
The flux limiter scheme has caused us lots of inexplicable slowdowns, and we were waiting with the merge until we resolved that along with the X-point temperature spike bug. With those two problems having turned out to be challenging, I think we should fix the equations in a separate PR as they are simply not correct. We can then tackle the limiter scheme methodically bit by bit.
I had to merge the other 2D neutral fixes PR into this as both are required for the tests. That other PR is causing checkerboarding in the neutral_mixed integrated test.
To do:
Original work from @bendudson:
Impact on results is not insubstantial, especially at the target. Performance is the same or slightly faster - probably dependent on the change in conditions rather than being more difficult numerically.
OMP
Target
Poloidal slice