-
Notifications
You must be signed in to change notification settings - Fork 0
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
FAIL: test_clustering_three_ensembles_two_identical on non-x86 64bit arches (aarch64, ppc64, ppc64le and s390x) #38
Comments
@mtiberti @wouterboomsma can you have a look at the test outputs here and give us some insights into how bad these failures in the ENCORE tests are? These are on architectures that we do not officially support but perhaps these failures point to some subtle bugs? |
For the record, this test is failing even on 32 bit x86 in 0.17.0:
|
@orbeckst Sorry, I missed your callout to me on this. We'll take a look at this immediately. |
Partially fixes #1389 and #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
Re-opening because we didn't address s390x. Is this something we want to target? We can add it to our Travis cron matrix, but do we actually have a reasonable expectation that any of our users are using IBM Z machines? As far as I'm aware there's not a single such machine in the top500, and I can't think of any academic cluster that has one either. Unless there's a reasonable case for it, I think it might be best to just say we don't cover it? |
…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
I'm closing this as a stale issue. Unless someone has a reasonable expectation that MDAnalysis needs to be run on IBM Z then supporting s390x really isn't necessary. If I'm wrong, please do re-open or open a new issue. |
This is not about s390x only. It's about test failures on four non-x86_64 arches and I'd argue at least aarch64 has become relevant now that Apple M1-based machines are out. I'll try to run a fresh build of the latest MDAnalysis source on all arches in Fedora and check what the current status is. |
I'm happy to re-open temporarily but MDA has minimal support (we can't guarantee optional dependencies) for non-M1 aarch64 and ppc64le since 2.0. Note: support for M1 is not happening until CI providers have M1 runners available. |
* the changes here allow me to build MDAnalysis and its dependencies on an M1 Mac on the gcc compile farm using `python3 -m pip install --user .` * the full test suite passes, except for 1 extra `xfail` I've added in here--which is also documented by gh-1389 (it has apparently been failing on "exotic platforms" for almost 5 years)
Partly addresses #3592 * BLD: improve M1 build support * the changes here allow me to build MDAnalysis and its dependencies on an M1 Mac on the gcc compile farm using `python3 -m pip install --user .` * the full test suite passes, except for 1 extra `xfail` I've added in here--which is also documented by gh-1389 (it has apparently been failing on "exotic platforms" for almost 5 years). Issue re-raised as #3599
Expected behaviour
All tests pass.
Actual behaviour
Code to reproduce the behaviour
On Fedora rawhide (27) running on aarch64, ppc64, ppc64le and s390x, the following failed test can be observed (log is from aarc64 build):
NOTE: on big-endian arches (ppc64 and s390x) there are also 67 test errors, but I'll open a separate issue for that.
Currently version of MDAnalysis:
0.16.1
The text was updated successfully, but these errors were encountered: