[mpiwg-tools] question about MPI_T_EVENT_SET_DROPPED_HANDLER
Raffenetti, Kenneth J.
raffenet at mcs.anl.gov
Mon Jan 25 15:07:16 CST 2021
On 1/25/21, 2:49 PM, "mpiwg-tools on behalf of Raffenetti, Kenneth J. via mpiwg-tools" <mpiwg-tools-bounces at lists.mpi-forum.org on behalf of mpiwg-tools at lists.mpi-forum.org> wrote:
On 1/25/21, 12:44 PM, "mpiwg-tools on behalf of Raffenetti, Kenneth J. via mpiwg-tools" <mpiwg-tools-bounces at lists.mpi-forum.org on behalf of mpiwg-tools at lists.mpi-forum.org> wrote:
On 1/25/21, 11:46 AM, "Marc-André Hermanns" <hermanns at itc.rwth-aachen.de> wrote:
For me (although not explicitly defined in the text) that would imply
that the dropped callback is only invoked once the library is in a
state where a callback (and user_data) for the same level is defined
and could be invoked.
Does this make sense and clarify your concerns? Do you think this need
more clarification in the text?
I see. Yes, that makes sense for my implementation issue.
I may have spoken too soon. I'm looking at pg 763, ln 37 (Handling Dropped Events):
If an event occurs...[and] no callback function meeting the required callback safety level is registered, the event data may be dropped. To discover such data loss, the user can set a handler function for a specific event-registration handle.
If a callback function meeting the required safety level is not registered, how do I decide which user buffer to provide to the dropped handler?
Oh, so if I am reading your clarification correctly above, you are saying I should not invoke the dropped handler until a callback for that safety level has been registered. It is possible then that I might never call the dropped handler. Should I be managing dropped event counts for all safety levels?
More information about the mpiwg-tools