(closes #3243) Unify NEMO parallelisation scripts#3443
Conversation
…ation' into single_nemo_script
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## master #3443 +/- ##
=======================================
Coverage 99.96% 99.96%
=======================================
Files 391 391
Lines 54689 54715 +26
=======================================
+ Hits 54670 54696 +26
Misses 19 19 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
|
@LonelyCat124 This is ready for another look, we had a first review some time ago in #3244 and there was open comments there. |
|
@LonelyCat124 Removing 9ee8eb6 has worked here, so this is ready to review again. |
There was a problem hiding this comment.
@sergisiso I think this is almost ready - have a few comments about the new README to fix up.
| > ``` | ||
| > when applying the transformations to NEMOv4. | ||
|
|
||
| TODO: Mention `ASYNC_PARALLEL`, `ENABLE_INLINING`, `PROFILING` |
There was a problem hiding this comment.
TODO - is this for this PR or for a future PR?
There was a problem hiding this comment.
I want to follow up this with with a few more changes but the PR has already become larger than I wanted, so I created a new #3445 issue to expand this and address other improvements. I added the number in the TODO tag.
| it is useful to apply the transformations step-by-step while checking if the | ||
| `run.stat` values change. Some useful steps are: | ||
|
|
||
| - Starting building NEMO *without* `psyclonefc` and conservative optimisation flags |
There was a problem hiding this comment.
Initially build NEMO ...
|
|
||
| - Starting building NEMO *without* `psyclonefc` and conservative optimisation flags | ||
| and run it serially (O2, no vectorisation, no-fma). Then store the generated `run.stat`. | ||
| - Then switch to using `psyclonefc` with the `PSYCLONE_OPTS="-s passthrough.py"`, |
There was a problem hiding this comment.
I'm not sure I like the repeated Then's here - could be better to use an enumerated list and remove the "Starting/Then/finall"?
| Alongside finding which step is causing the divergence we may want to find | ||
| which file/s are causing it. This folder also contains a `do_file_by_file.sh` | ||
| script that build NEMO many times, each with only one file being transformed, | ||
| and compares the results with the stores `run.stat` |
There was a problem hiding this comment.
Done, but I also commented this paragraph and added a TODO to the follow up PR as I haven't added the do_file_by_file.sh script yet.
|
@LonelyCat124 This is ready for another review |
2 participants
This is the continuation of #3244 , I started a new PR because I shared the branch associated with the previous PR with multiple people that may be using them and modifying may break their workflows.
The PR aims to: