PRRTE v3.0 and other matters
6 views
Skip to first unread message
Ralph Castain
Jul 1, 2022, 4:05:40 PM7/1/22
to 'Thomas Naughton' via pmix
Hi folks
I've been working on a major revision/update to the PRRTE map/rank/bind system and am close to committing it. This will remove one of the major blockers for the next PRRTE release, leaving only two more to complete. In doing this work, it has become apparent that there is no way to justify doing this next release as a member of the PRRTE v2 series. There is simply too much change in it.
Accordingly, I will be deleting the PRRTE v2.1 branch and replacing it with a new PRRTE v3.0 release branch based on the current master branch. This will happen towards the middle of July (after the map/rank/bind update) - no point in doing it earlier and having to cherry-pick more stuff across.
PRRTE v3 will depend on a minimum of PMIx v4.2 to build/operate. I had contemplated supporting it with PMIx v4.1.3, but the changes required really are inappropriate for a bug-fix release. I'll back the v4.1 branch down to the v4.1.2 level and look at potentially building a v4.1.3 release from true bug fixes.
I'll send out more information on the map/rank/bind changes separately. They shouldn't impact the majority of users out there, but they allow the map/rank/bind operation to be conducted much faster than before, and provide a more mantainable code base.
Ralph
I've been working on a major revision/update to the PRRTE map/rank/bind system and am close to committing it. This will remove one of the major blockers for the next PRRTE release, leaving only two more to complete. In doing this work, it has become apparent that there is no way to justify doing this next release as a member of the PRRTE v2 series. There is simply too much change in it.
Accordingly, I will be deleting the PRRTE v2.1 branch and replacing it with a new PRRTE v3.0 release branch based on the current master branch. This will happen towards the middle of July (after the map/rank/bind update) - no point in doing it earlier and having to cherry-pick more stuff across.
PRRTE v3 will depend on a minimum of PMIx v4.2 to build/operate. I had contemplated supporting it with PMIx v4.1.3, but the changes required really are inappropriate for a bug-fix release. I'll back the v4.1 branch down to the v4.1.2 level and look at potentially building a v4.1.3 release from true bug fixes.
I'll send out more information on the map/rank/bind changes separately. They shouldn't impact the majority of users out there, but they allow the map/rank/bind operation to be conducted much faster than before, and provide a more mantainable code base.
Ralph
Reply all
Reply to author
Forward
0 new messages