Update GEOSadas to be in line with GEOSgcm v10.19.3 #108
Conversation
|
Yes, as Matt said with the ops emissions there was a small non-zero diff. The issue was that I had previously made a change to ExtData that caused the order the variable to be provided get processed to change. Well it turned out there were two emissions files that were identified as having "equal" lat-lon grids but on close inspection they have ever so slightly different lons and lats (i'm talking very small differences). Well, which ever file got opened first, that lat-lon grid won our grid manger so when the next file was opened it saw, I've already created a grid with that factory. So depending on which was opened first you would get slightly different lons and lats. So this change I made caused this order of these to flip during processing and indeed just change the order these two entries were in the ExtData file would change which got used. So when we realized this we tightened up the lat-lon grid factory so inconsistencies would not happen like this again. |
|
Call me crazy ... but I can no longer find the repo for fvdycore! What happened to it? I am trying to see the diff between what I have in develop (geos/v1.1.6) and what you have here ... but I can find the repo! |
Ah yeah, that was one where the actual repo has a really different name than what GEOS called it. NOAA calls it "GFDL_atmos_cubed_sphere": https://github.com/GEOS-ESM/GFDL_atmos_cubed_sphere So the difference would be: GEOS-ESM/GFDL_atmos_cubed_sphere@geos/v1.1.6...GEOS-ESM:geos/v1.1.7 The reason for this was documented by @bena-nasa here: GEOS-ESM/GFDL_atmos_cubed_sphere#43 |
|
It's really odd how and why an optional variable needs to be ifdef'ed around it! Anyway, I will assume this is as the change claims - to be zero-diff. Thanks Matt. |
|
Another odd event in which I get: git checkout v1.1.7 What's the trick here? |
The tag is actuall |
There was a problem hiding this comment.
I regret to say, but these are not coming up zero-diff to me. I have not looked closely enough to find out where the diff comes from, but I can say the diffs from the DAS are only in the model and, after 12 hours of run they look like roundoff - no clear patterns. Still, I am reluctant to take the combination.
|
@rtodling The summary at the top definitely states that the differences are not always 0-diff. Probably should have made the statements in the reverse order to minimize communication. And we should have noticed that your Aug. 3rd comment suggest you had missed the non-zero diff aspect. |
|
Now included in GEOSadas-5_29_3: MAPL v2.8.0 As for Shared ... I need feature/rtodling/cascase for the DAS; I think the DAS supersede the version pointed to here: v1.4.6 As for GEOSgcm_App: The DAS uses what's under: feature/rtodling/v1_5_4_onlXoffl_flags which (as far as I can tell) supersedes v1.5.3 |
|
@rtodling Indeed, this is probably out of date. I'll close this as your branches are probably above this. |
In the time since #103 was made, the GCM updated to v10.19.3. This PR would update the GEOSadas as well. The updates are:
GMAO_Shared v1.4.6
MAPL v2.8.0
GEOSgcm_GridComp v1.12.4
fvdycore v1.1.7
GEOSgcm_App v1.5.3
Note that this is potentially non-zero-diff. Per the GEOSgcm release notes:
@bena-nasa can explain the MAPL bug fix better than I can if needed.