(closes #3475) Make OMPParallelLoopTrans respect the force_private kwarg#3476
Conversation
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## master #3476 +/- ##
=========================================
Coverage 100.00% 100.00%
=========================================
Files 395 395
Lines 55199 55215 +16
=========================================
+ Hits 55199 55215 +16 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
There was a problem hiding this comment.
Ah I think this is why this escaped my updates before. This transformation just force overrides specified options even if they've been provided. I think we can make this nicer later - essentially now that we have inherit=False working, we can avoid exposing the reduction_ops kwarg in the docstring, or directly on the function. This means we can always just ignore it (perhaps with a warning if reduction_ops is in **kwargs)
The options.copy() is just because the transformation can modify the dictionary and this needs to be modified, but as you say we don't need to fix this now.
I'll build the docs and check those quick, but otherwise this is ready to merge.
The only other thing is that this transformation currently won't match rules w.r.t options vs kwargs, but again not essential right now but we should fix as soon as possible after.
@sergisiso I assume this one needs merging now before the release?
|
All is ready to be merged if the CI passes. |
This fixes one of the issues that LFRic_apps is having to adopt psyclone 3.3.
@LonelyCat124 I got a bit nervous about modifying the existing 'enable_reductions' option because it does a local
options.copy()that I don't understand. So for the time being I left it as an option. I propose we unify this after the release.