-
Notifications
You must be signed in to change notification settings - Fork 934
ompi: ompi_mpi_init(): do not export threading level to modex. #4837
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
Conversation
For some of our configuration this flag increases per-process contribution by ~20% while it is not being used currently. The consumer of this flag was communicator ID calculation logic, but it was changed in 0bf06de. Signed-off-by: Artem Polyakov <artpol84@gmail.com> (cherry picked from commit b601dd5)
rhc54
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I didn't find a MODEX_RECV for MPI_THREAD_LEVEL, but I did find:
ompi/mpi/c/init.c
54: if (NULL != (env = getenv("OMPI_MPI_THREAD_LEVEL"))) {
Does that need to be removed? I couldn't find where it gets set.
|
@rhc54 unfortunately I don't have a spare cycle to investigate that. And I don't want to change it blindly. |
|
@bwbarrett @jjhursey I'll have to defer to someone else to investigate, then - I don't know if somehow these things are tied together and your tests simply aren't tripping over it. Could be someone out there has an idea? |
|
@rhc54 if you can't find modex recv, I think this means that presence of this information in the modex doesn't play a role. Isn't it? |
|
bot:ompi:retest For some reason, the OMPI tests tried to pull a commit that doesn't exist. Very weird. |
|
I mean that rank don't need to care about someone else's threading level, but it still can be concerned about its own threading level which might be exported through env. |
|
My recollection is that OMPI_MPI_THREAD_LEVEL allows to start the MPI library in full threading mode even if the user calls MPI_Init. According to our 2015 developer meeting notes we were using it to test for correctness on all nightly tests (at least at Cisco and LANL). |
|
@bosilca Appreciate the explanation - I couldn't see where it was being set, nor if it was somehow related to the exchange. Sounds like it is unrelated. |
|
@bosilca indeed thanks for the explanation! |
For some of our configuration this flag increases per-process contribution
by ~20% while it is not being used currently.
The consumer of this flag was communicator ID calculation logic, but it was
changed in 0bf06de.
Signed-off-by: Artem Polyakov artpol84@gmail.com
(cherry picked from commit b601dd5)