-
Notifications
You must be signed in to change notification settings - Fork 15
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add 0-moment micro to diagnostic EDMFX #2011
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looking good. I was hoping we can have something like calculate_qt_tendency
and dispatch on the microphysics/precipitation scheme. And for now we can have nothing or 0M, so we can still support e.g. Bomex where we don't have precipitation. And we can add 1M in the future.
e6b4cb2
to
8f7fc1e
Compare
389053b
to
1460e82
Compare
What's the advantage of caching |
Initially I thought I would be able to hide the different options via dispatch over So from that point of view I can instead cache |
All the cases work now. From the precipitating ones:
I'll tackle allocations tomorrow, though right now I don't see any obvious culprits |
5aab901
to
e610a5b
Compare
@charleskawczynski - I applied your suggestions regarding the way I was summing things in precipitation tendency and with the FT. However it's still allocating. Is that level of allcocs acceptable? Or would you have any more hints for me?
|
FYI: Just adding 3 variables to cache
and then unpacking them:
Increases allocations by
See this PR: #2031 I'll test now just adding them, without unpacking. |
Interesting, I wonder if we're running into CliMA/ClimaCore.jl#1015? |
I think it might be better / more ideal if we could aim to compute more things on the fly, from a minimal number of variables, if possible. Otherwise, it's probably fine to just increase the allocation limit so that you're not held back by this. |
Yes, I agree. Unfortunately we have to cache those ones. Otherwise we would have to redo the whole EDMF diagnostic loop to get the SGS contributions. I'll test if it has something to do with unpacking rather than caching. But I doubt it |
What we should be doing to avoid caching and allocating, is to augment the grid-mean tendencies from within the Maybe we could go back to it once we delete the old EDMF code? It will be easier to make the changes then. This PR only increases the allocations by |
c52c7f4
to
e53b5c4
Compare
e53b5c4
to
2c4cdb8
Compare
bors r+ |
Build succeeded! The publicly hosted instance of bors-ng is deprecated and will go away soon. If you want to self-host your own instance, instructions are here. If you want to switch to GitHub's built-in merge queue, visit their help page. |
This PR couples 0-moment microphysics scheme to diagnostic EDMFX