You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Turns out the solution to #710 introduced the following text into current MPI 4.1 standard in the discussion of MPI_Session_finalize, namely
Most handles that exist independently from theWorld Model or a session in the Sessions
Model, e.g., datatype handles, can be created only while MPI is initialized. For example,
a datatype handle that was created when one particular session existed can be used in
any other session (or in the World Model), even if the second session was initialized after
the first session had already been finalized and no other session existed in between. See
Section 11.4.1 for handle creation procedures that do not require that MPI is initialized.
The problematic part of this text is the statement is even if the second session was initialized after the first session had already been finalized and no other session existed in between them. Its not clear to implementors how to implement this.
This was reflected in the voting results for the solution to #710 at the last MPI forum.
Proposal
TBC
Changes to the Text
TBC
Impact on Implementations
Potentially significant. A simple ref counting of MPI sessions (counting world process model as one session) would no longer suffice for components of an MPI implementation responsible for managing these types of handles.
Impact on Users
Unclear due to limited use of the sessions model currently
References and Pull Requests
The text was updated successfully, but these errors were encountered:
In the Sessions WG there was discussion of new functions taking a sessions handle for datatype constructor functions, etc. so it looks like a good solution is beyond 4.1.
Problem
Turns out the solution to #710 introduced the following text into current MPI 4.1 standard in the discussion of
MPI_Session_finalize
, namelyThe problematic part of this text is the statement is
even if the second session was initialized after the first session had already been finalized and no other session existed in between them
. Its not clear to implementors how to implement this.This was reflected in the voting results for the solution to #710 at the last MPI forum.
Proposal
TBC
Changes to the Text
TBC
Impact on Implementations
Potentially significant. A simple ref counting of MPI sessions (counting world process model as one session) would no longer suffice for components of an MPI implementation responsible for managing these types of handles.
Impact on Users
Unclear due to limited use of the sessions model currently
References and Pull Requests
The text was updated successfully, but these errors were encountered: