Closed timfelle closed 2 weeks ago
As a first step, we need to document the fields that are generated and put into the s1... fields in the .flds.
I guess we need to talk more about the statistics in general. I know we did that already a few times, but the strategy needs to be clear. It is a major effort, but crucially important I would say. My recommendation is in fact to start with a prototype of just first and second order statistics, to explore how it can be done.
As a first step, we need to document the fields that are generated and put into the s1... fields in the .flds.
My thoughts exactly. And reading the code was not enough for me to get a grip on it and start a draft PR.
In general it would be great if the fld files could contain the names of the fields, as the field_t
already contain the name.
Yeah, hdf5 will hopefully solver the latter in the future
All of these are good points I think, when I added stats originally I mostly wanted to make sure that all data that could be relevant in a sim was dumped to file, but the documentation is really lacking.
To start, I will add a list of the fields that are stored in the stats file to better document what is there currently. Then we can also look into user-cusomizable stats and so on as you say Philipp, but to start Ill just document what is there.
Created an initial draft.
As evident from the discussion https://github.com/ExtremeFLOW/neko/discussions/1130, currently the statistics are poorly documented. This needs to be adjusted, hopefully we can make it before 0.8.