Cubed sphere interpolation #90
Conversation
Codecov Report
@@ Coverage Diff @@
## feature/cubed-sphere-base3 #90 +/- ##
==============================================================
- Coverage 78.01% 77.93% -0.08%
==============================================================
Files 773 781 +8
Lines 53089 53346 +257
==============================================================
+ Hits 41418 41577 +159
- Misses 11671 11769 +98
Continue to review full report at Codecov.
|
There was a problem hiding this comment.
I don't think this will cause any issues with the changes I have been making.
There is a chance you are missing out on some small optimisations in localRemap with your refactor. I am not sure if it is worth looking at, but as part of the response to review of my PR for this I removed the Roots struct and calculated the weight directly. I really don't know but it is possible that vectorises better? Having said that, my brief test on a ~1million obs seems to run in about the same time so I am not concerned.
#96) * Use the 3D kdtree to locate grid elements * retry observations which fail in case they are in a cell which * overlaps the eastern edge of the grid
|
Closed due to deprecated base branch. |
Note: this PR cannot go into ecmwf/develop until PR ecmwf#95 passes review. feature/cubed-sphere-base3 is ecmwf/develop with ecmwf#95 merged into it.
This feature adds following functionality:
CS-LFR-24 interpolation source field:
O48 interpolation target field:
@twsearle Can you check that my changes to Quad2D and Triag2D don't break anything you're currently working on. The ctests appear to run fine.
@wdeconinck and @MarekWlasak I basically copy-and-pasted code from StructuredColumns to NodeColumns to get the adjoint of the halo exchange to work. Can you check that I haven't done something truly daft?
closes #82 and #83