Skip to content

Initialize opal/smsc outside of btl/sm, to enable its use without it #10897

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

Merged
merged 1 commit into from
Oct 12, 2022

Conversation

gkatev
Copy link
Contributor

@gkatev gkatev commented Oct 6, 2022

Hello, this places the initialization call for opal/smsc in an allegedly (more) appropriate place -- I'm accepting input about the "legality" of its placement :-). Currently, the call is placed inside btl/sm, but if it is not loaded (e.g. because pml=ucx), smsc will remain uninitialized. My initial thought was that a component requiring smsc could initialize it on demand as necessary, but this might be too late for the modex to be properly sent/commited. For example, this occurs with collectives components that utilize smsc (#10342).

Fixes #10342
Signed-off-by: George Katevenis [email protected]

@ompiteam-bot
Copy link

Can one of the admins verify this patch?

Copy link
Contributor

@devreal devreal left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@awlauria awlauria merged commit 6c90ecc into open-mpi:main Oct 12, 2022
@awlauria
Copy link
Contributor

Please bring back to v5.0.x

@@ -498,6 +498,10 @@ static int ompi_mpi_instance_init_common (int argc, char **argv)

/* Select which MPI components to use */

if (OPAL_SUCCESS != (ret = mca_smsc_base_select())) {
return ompi_instance_print_error ("mca_smsc_base_select() failed", ret);
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Previously, OMPI would silently handle smsc not being initialized. Now it's printing a warning. That does not seem like the behavior we want?

Copy link
Contributor Author

@gkatev gkatev Oct 12, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

At the moment mca_smsc_base_select() only returns OPAL_SUCCESS. The actual smsc components might still remain silently uninitialized. My though process here is that if it at some point in the future mca_smsc_base_select() started possibly returning something else, it would be something severe enough to warrant triggering an error.

@gkatev gkatev deleted the smsc_fix branch October 17, 2022 09:04
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

How to initialize smsc when btl/sm is not used?
5 participants