Closed koomie closed 3 years ago
This is nice since we'll be able to see the evolution of the mean as the simulation marches on. (This will output an average solution for each cycle). Do we want to remove then the folder with the average solution?
This is nice since we'll be able to see the evolution of the mean as the simulation marches on. (This will output an average solution for each cycle). Do we want to remove then the folder with the average solution?
You mean the location in HDF5? I'd like to keep it there as well, but wanted to add to paraview for now so we can access it easily.
But there is already paraview output. It's in a different folder called mean_{output_name}
But there is already paraview output. It's in a different folder called
mean_{output_name}
Doh. I completely missed that. I guess I thought it would have been more convenient to keep everything in one file so we don't have to write and load the mesh twice.
It doesn't look like you are setting the iteration and time for your mean_output
(using SetCycle
and SetTime
). Doesn't that mean it's going to get overwritten every time? I'd like to be able to keep these the same way we keep current primitive outputs for paraview output so we can inspect easily at different times.
Seems like a quick fix would be to update Averaging::write_meanANDrms_restart_files()
to set the matching time and iteration so it is consistent with the paraviewColl
primitive output (in which case, this current PR can be ignored). Or we could put everything in the same paraview file.
Preferences?
It is nice to see the evolution of the mean with time though it comes at the price of having solution files that are more than 3 times the size (except for the mesh). This is the reason why I don't set the iteration. That way the files are overwritten and the solution is lighter.
In order to check the evolution of the mean, we could add an option in Averaging::write_meanANDrms_restart_files()
such that it sets the cycle and time that we can activate and deactivate. That way we can have both: save space when we want and have evolution when we want to check that averaging is moving in the right direction.
Another (cheap) option is related to what we talked about and Raphael commented last Monday: output domain-integrated variables in a text file. We could do that for both the instantaneous solution (so we know when to start averaging) and with averaged solution, so that we know when the mean is converged.
In order to check the evolution of the mean, we could add an option in
Averaging::write_meanANDrms_restart_files()
such that it sets the cycle and time that we can activate and deactivate. That way we can have both: save space when we want and have evolution when we want to check that averaging is moving in the right direction.Yes. We need this, otherwise we will have to remember to cache copies of
mean_output
by hand before restarting. Can you please add this? And then, I'll just toss this PR.Another (cheap) option is related to what we talked about and Raphael commented last Monday: output domain-integrated variables in a text file. We could do that for both the instantaneous solution (so we know when to start averaging) and with averaged solution, so that we know when the mean is converged.
Yep - need this too.
See what you think of #53 for the first part discussed above to add an option to allow for the mean solution history to be saved.
See what you think of #53 for the first part discussed above to add an option to allow for the mean solution history to be saved.
I like it. Do you want to merge #53 and I get on with the domain-integrated output?
That sounds good. I just landed #53.
That sounds good. I just landed #53.
We can close this PR then
… enabled