[Mpi-forum] Voting results
jsquyres at cisco.com
Wed May 30 13:36:09 CDT 2012
Voting results from the May 2012 Japan MPI Forum meeting are now up:
1. 3 tickets were withdrawn:
Ensure MPI_Dims_create is really suitable for MPI_Cart_create
MPI3 Fault Tolerance - Files
MPI3 Fault Tolerance - Dynamic process management
2. The definition of "simple majority" was changed from how I have computed whether ballots passed or failed in the past. I don't know offhand how past ballot results would have fared with the new definition; I am guessing that their results wouldn't have changed because most past ballots were not as close as some of the ones from this week.
>From my understanding, "simple majority" (i.e., what a vote needs to pass) was defined as the following:
floor(total_eligible_orgs_attending / 2) + 1 "yes" votes
Meaning: abstains and misses count as "not yes", or (effectively) "no".
*** With these rules, I see no meaning for "abstain" (or "miss"). There is effectively only "yes" and "no".
*** Meaning: everyone who thought they were abstaining at this past meeting were actually voting "no".
I understand that this was discussed in Japan and everyone in the room agreed to these rules. ***It is not what I would have advocated***, but I was not there. :-\
In all prior meetings, I used the following computation to determine if a ballot passed:
floor(total_yes_and_no_votes / 2) + 1 "yes" votes
more "yes" votes than "no" votes
Meaning: abstains and misses do not count towards the result.
3. 16 organizations were eligible to vote at this meeting.
4. In this meeting, 12 first votes passed. 8 of those 12 had good consensus (max of 3 (no+abstain)), but 4 did not:
- #278 (fix deprecated examples): 6 abstains
- #281 (remove C++): 5 no votes
- #303 (removed interface chapter): 5 no votes, 2 abstains
- #313 (INIT/FINALIZE): 1 no vote, 4 abstains
5. In this meeting, 5 first votes failed:
Topology awareness in MPI_Dims_create
MPI3 Hybrid Programming: Proposal for Helper Threads
Add Immediate versions of nonblocking collective I/O routines
clarify MPI behavior when multiple MPI processes run in the same address space
User-Level Failure Mitigation
6. In this meeting, 8 second votes passed (i.e., all of them). Good consensus on all of them (max of 3 (no+abstain) votes).
jsquyres at cisco.com
For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/
More information about the mpi-forum