Hello,
as a user I would like to give the following suggestion. I find annoying to stop a simulation and restart it to make the fields averages whenever I converge to some steady-on-average conditions, so why not putting an option in the control file that makes the averaging start whenever I want?
In the end it would be necessary just to edit dynamically the ntdmom(imom) or ttdmom(imom) variables and keep everything else as before.
Would it be feasible?
Cheers!
In-run average setting
-
Yvan Fournier
- Posts: 4080
- Joined: Mon Feb 20, 2012 3:25 pm
Re: In-run average setting
Hello Jacopo,
It would be relatively easy to adapt the control_file to reset the starting time for defined time moments. It would be a bit more complicated to decide that you want an average if none was declared before.
At some point though, we may need to decide which variables may be dynamically changed, and which may not, and document that. It is already possible for a user to test every n time steps in cs_user_extra_operations.f90 (or oher user subroutines) for the existence of a file of his own, and read and change options there, so a "poor man's computational steering" is already possible, but having something easy to use and well documented needs some thought. This would be interesting, though, to avoid option compatibility issues (i.e. you may change the reference time step during the calculation, but not the turbulence model).
Cheers,
Yvan
It would be relatively easy to adapt the control_file to reset the starting time for defined time moments. It would be a bit more complicated to decide that you want an average if none was declared before.
At some point though, we may need to decide which variables may be dynamically changed, and which may not, and document that. It is already possible for a user to test every n time steps in cs_user_extra_operations.f90 (or oher user subroutines) for the existence of a file of his own, and read and change options there, so a "poor man's computational steering" is already possible, but having something easy to use and well documented needs some thought. This would be interesting, though, to avoid option compatibility issues (i.e. you may change the reference time step during the calculation, but not the turbulence model).
Cheers,
Yvan