Skip to content

README: update to indicate aarch64 is not currently supported #3855

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

Closed
hppritcha opened this issue Jul 12, 2017 · 5 comments
Closed

README: update to indicate aarch64 is not currently supported #3855

hppritcha opened this issue Jul 12, 2017 · 5 comments
Assignees
Milestone

Comments

@hppritcha
Copy link
Member

The Open MPI project currently lacks resources/interest to actively support aarch64. This needs to be updated in the README.

@hppritcha hppritcha added this to the v3.0.0 milestone Jul 12, 2017
@hppritcha hppritcha self-assigned this Jul 12, 2017
hppritcha added a commit to hppritcha/ompi that referenced this issue Jul 12, 2017
@ggouaillardet
Copy link
Contributor

@hppritcha fwiw, i build Open MPI on my raspberry pi 3 running Open SuSE Leap 42.2 once in a while, and so far, i have no issue with it (i test is very lightly though).
i am concerned stating bluntly "Open MPI does not support ARM 64bits" is both strong and could be misleading (some vendors have such plans, @kawashima-fj can you confirm post-K (Fujitsu/Riken co-designed ARM64 based processor) will use Open MPI ?)

@jsquyres
Copy link
Member

#3697 (comment) is a good summary of the current status of ARM v8 support in OMPI v2.1.x (I assume it's similar in v2.0.x) and 3.0.0rc1. The consensus in the room yesterday is that if ARM is going to move to be a supported platform, we need regular testing results in MTT.

@bwbarrett
Copy link
Member

Yes, but @ggouaillardet raises a good point. We don't actually document what's not supported. We document what's supported and what's known not to work. It's a subtle distinction, but one we should follow here. aarch64 is already not listed as a supported platform. What we need to do is document the known issues with the platform, so that others aren't surprised by them.

@shamisp
Copy link
Contributor

shamisp commented Jul 13, 2017

Description in #2643 (comment)
is not accurate.

We have been responding on both issue (reflected in tickets). The issues are not easily reproducible since it depends on multiple factors, including ARMv8 implementation (multiple vendors), drivers, compiler, etc. In some case we directly approached people for remote access, but we understand it is not always possible.

Besides this, some of these problem not ARM specific and have been reported on other platforms, like PPC. Moreover neither of issue was reported as a blocker.

In terms of MTT - we are working to enable this and this takes time. In addition, we proactively added OMPI Jenkins build node with IB to HPCAC center for OpenMPI testing in public domain. As for now, it has some stability issues (related to OS) but we have been working to address this.

More reasonable approach is to list some of the know issue in know bug section (including precise platform description, since there are multiple ARMv8 implementations).

@kawashima-fj
Copy link
Member

@ggouaillardet The post-K computer will use Fujitsu ARMv8 CPU and Fujitsu has great interest in supporting ARMv8. But our team lacks machine resource to run MTT currently.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

6 participants