Skip to content
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

MPI_Group_from_session_pset should use PMIX_QUERY_PSET_MEMBERSHIP #10862

Closed
hppritcha opened this issue Sep 28, 2022 · 4 comments
Closed

MPI_Group_from_session_pset should use PMIX_QUERY_PSET_MEMBERSHIP #10862

hppritcha opened this issue Sep 28, 2022 · 4 comments
Assignees

Comments

@hppritcha
Copy link
Member

The current implementation of MPI_Group_from_session_pset will only work for process sets defined as part of the job launch mechanism. PMIx allows for a more general approach for querying for the list of pmix_proc's in a pset. We should use this query function to support use cases involving more flexible generation of pset names going forward.

@hppritcha hppritcha self-assigned this Sep 28, 2022
@rhc54
Copy link
Contributor

rhc54 commented Sep 28, 2022

FWIW: IIRC, I added the ability for PMIx to dynamically define process sets. I can check and ensure PRRTE supports it as well. Not sure if that is relevant to what you are denoting here, but thought I'd pass it slong just in case.

@rhc54
Copy link
Contributor

rhc54 commented Sep 28, 2022

Also FWIW - you are correct about using the query function. Only the initial process set definitions are available to all procs in the job info. Dynamically created ones are known to their participants - others need to query for them. I'll ensure PRRTE supports that query.

hppritcha added a commit to hppritcha/ompi that referenced this issue Oct 3, 2022
It turns out that the existing ompi_instance_group_pmix_pset
implementation assumes an MPI_COMM_WORLD type of model.

This prevents the ability to use more dynamically generated process
sets, possibly using an external agent.

Swith to using the pmix pset membership query to find new pset
membership.

Related to open-mpi#10862

Signed-off-by: Howard Pritchard <howardp@lanl.gov>
hppritcha added a commit to hppritcha/ompi that referenced this issue Oct 4, 2022
It turns out that the existing ompi_instance_group_pmix_pset
implementation assumes an MPI_COMM_WORLD type of model.

This prevents the ability to use more dynamically generated process
sets, possibly using an external agent.

Swith to using the pmix pset membership query to find new pset
membership.

Related to open-mpi#10862

Signed-off-by: Howard Pritchard <howardp@lanl.gov>
hppritcha added a commit to hppritcha/ompi that referenced this issue Jan 9, 2024
from PMIx.

Related to open-mpi#10862

Signed-off-by: Howard Pritchard <howardp@lanl.gov>
hppritcha added a commit to hppritcha/prrte that referenced this issue Jan 10, 2024
query.

Related to open-mpi/ompi#10862

Signed-off-by: Howard Pritchard <howardp@lanl.gov>
hppritcha added a commit to hppritcha/ompi that referenced this issue Jan 10, 2024
It turns out that the existing ompi_instance_group_pmix_pset
implementation assumes an MPI_COMM_WORLD type of model.

This prevents the ability to use more dynamically generated process
sets, possibly using an external agent.

Switch to using the pmix pset membership query to find new pset
membership.

Related to open-mpi#10862
Related to openpmix/prrte#1906

prrte changes in above referenced PR are necessary for creating
groups/communicators from psets defined by --pset option on the
mpirun command line.

Signed-off-by: Howard Pritchard <howardp@lanl.gov>
rhc54 pushed a commit to openpmix/prrte that referenced this issue Jan 11, 2024
query.

Related to open-mpi/ompi#10862

Signed-off-by: Howard Pritchard <howardp@lanl.gov>
@hppritcha
Copy link
Member Author

bot:aws:retest

rhc54 pushed a commit to rhc54/prrte that referenced this issue Jan 11, 2024
query.

Related to open-mpi/ompi#10862

Signed-off-by: Howard Pritchard <howardp@lanl.gov>
(cherry picked from commit 5cb219c)
rhc54 pushed a commit to openpmix/prrte that referenced this issue Jan 11, 2024
query.

Related to open-mpi/ompi#10862

Signed-off-by: Howard Pritchard <howardp@lanl.gov>
(cherry picked from commit 5cb219c)
hppritcha added a commit to hppritcha/ompi that referenced this issue Jan 11, 2024
It turns out that the existing ompi_instance_group_pmix_pset
implementation assumes an MPI_COMM_WORLD type of model.

This prevents the ability to use more dynamically generated process
sets, possibly using an external agent.

Switch to using the pmix pset membership query to find new pset
membership.

Related to open-mpi#10862
Related to openpmix/prrte#1906

prrte changes in above referenced PR are necessary for creating
groups/communicators from psets defined by --pset option on the
mpirun command line.

Signed-off-by: Howard Pritchard <howardp@lanl.gov>
@hppritcha
Copy link
Member Author

fixed via multiple commits both in openmpi and prrte

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

No branches or pull requests

2 participants