- changed milestone to 8.1
vnmult saved in restart file but not restored on restart
In the run_parameters
code handling setting the time step from the restart file when delt_option = 'check_restart'
we also try to restore the vnmult
values used by collisions. Unfortunately we don't actually do anything with these values currently. We should ensure that collisions:vnmult
is set from the data that's read in.
Comments (5)
-
reporter -
reporter There could be some other small bugs associated with this. Currently we set vnmult to 1 by default and this only becomes different from 1 if we call get_verr and
vary_vnew
is active (and we think increasing the collision frequency will help with errors). If we do set vnmult to something other than 1 we will then reset this to 0 during a time step change (due tovnmult=0
infinish_collisions
) before setting it back to 1 in our usual call toinit_lorentz/init_ediffuse
as a part of re-initialising collisions. This means our effective collision frequency will drop back to the input value value in the period between a timestep change and the next call toget_verr
. -
reporter PR #530 should hopefully fix the above issues.
In making this fix I realised that when we change
vnmult
during the time advance (through calls toget_verr
) we are effectively modifying the collision operator. Strictly we might expect to have to recalculate the response matrices to account for this change in our implicit solve. It might be worth exploring the potential error that this introduces. -
reporter The above should be PR #531
-
reporter - changed status to resolved
Fixed in 8.1 through PR #531.
Note additional potential issue flagged here.
- Log in to comment