the one degree simulation in main is quite weird with temperatures going a bit wild. https://numericalearth.github.io/NumericalEarthDocumentation/dev/literated/one_degree_simulation/
This is most likely caused by changing to zstar and I think the problems might be related to one of the two:
- radiative forcing (I am looking into it, seems to increase the surface temperature a lot with zstar)
- GM (uses horizontal derivatives that might interpolate in dead cells)
Also the performance is quite wack ~ 2.6 minutes with zstar vs 1.5 minutes without zstar (dev vs stable), about 70% slower! zstar was typically 10% slower but with 70% starts to become unusable. It is most likely caused by the horizontal derivatives (in GM and biharmonic viscosity) interpolating in z making the tracer tendencies complicated and register hungry. I will look into it more.
the one degree simulation in main is quite weird with temperatures going a bit wild. https://numericalearth.github.io/NumericalEarthDocumentation/dev/literated/one_degree_simulation/
This is most likely caused by changing to zstar and I think the problems might be related to one of the two:
Also the performance is quite wack ~ 2.6 minutes with zstar vs 1.5 minutes without zstar (dev vs stable), about 70% slower! zstar was typically 10% slower but with 70% starts to become unusable. It is most likely caused by the horizontal derivatives (in GM and biharmonic viscosity) interpolating in z making the tracer tendencies complicated and register hungry. I will look into it more.