[MAINT] Trying to re-enable testing for non-x86 archs #3149
Conversation
Codecov Report
@@ Coverage Diff @@
## develop MDAnalysis/mdanalysis#3149 +/- ##
===========================================
- Coverage 92.68% 92.68% -0.01%
===========================================
Files 171 168 -3
Lines 22737 22677 -60
Branches 3212 3212
===========================================
- Hits 21074 21018 -56
+ Misses 1615 1611 -4
Partials 48 48 Continue to review full report at Codecov.
|
|
closing & re-opening due to some GH issues that have stalled updating my latest commits. |
|
That's weird, looks like conda is picking up an old version of scipy even thought numpy is picked up as 1.20. I can pin CI to >1.5 I guess. Any thoughts @tylerjereddy ? https://travis-ci.com/github/MDAnalysis/mdanalysis/jobs/490647828#L2460 |
|
Hmm, one difference is that SciPy is listed explicitly while NumPy is not (presumably implicit), for the I wouldn't think that would be the cause though; the ppc64le arch is certainly not as well supported, but there seems to be an appropriate conda binary for modern SciPy on that platform here: https://repo.anaconda.com/pkgs/main/linux-ppc64le/ |
Strange it does pick it up if told to (or afterwards if doing it interactively). I won't question it too much, we can always pin the versions when we release on conda-forge. |
|
The provenance of 98 new user warnings is baffling to me: https://github.com/MDAnalysis/mdanalysis/pull/3149/checks?check_run_id=2099327091#step:10:317 |
| name: "Power PC" | ||
| os: linux | ||
| arch: ppc64le | ||
| if: type = cron |
There was a problem hiding this comment.
I'll be honest, I'm not 100% sure this will work. I have enabled the cron job on develop to be daily for now (eventually we can change it to weekly). So maybe it'll get triggered tomorrow, but it's not clear to me that the cron job will run if this doesn't get merged first.
| - python: 3.8 | ||
| name: "Power PC" | ||
| os: linux | ||
| arch: ppc64le |
There was a problem hiding this comment.
We can probably add s390x here for the sake of #1389, but is IBM Z really that prevalent? I guess the z15 was released in 2019, but I can't think of a single cluster that uses it? (definitely nothing in the top500 list)
|
Cheers @richardjgowers I'll merge so we can test that today's cron job is working as intended. We can discuss s390x further in MDAnalysis/mdaencore#38 |
…lysis#3149) Partially fixes #1389 and MDAnalysis#3127 # Work done in this PR - Adds basic support for ppc64le - Adds encore specific C compiler args to reduce optimisations in non-x86 systems - Enables travis cron CI for aarch64 and ppc64le
Partially fixes MDAnalysis/mdaencore#38 and #3127
Changes made in this Pull Request:
PR Checklist