eigen3 is consumed by pybind11, which is consumed by scipy (CCing Matej Muzila and Nikola Forro).

Tom, does your team has the capacity to handle the Eigen3 3.3.9 -> 3.4 rebase in 9.0 GA or it is too late?

(In reply to Jiri Kucera from comment #2)
> Tom, does your team has the capacity to handle the Eigen3 3.3.9 -> 3.4
> rebase in 9.0 GA or it is too late?

Hello Jirko,
unfortunmately we do not plan accept more components in rhel9 and this one seems to be a leftover from our perspective. There is no other desktop pkg that would be dependen on this pkg. As you mentioned in c1 only pybind11. It would also make sense if we move Eigen3 under team which actually use it.

------- Comment From fnovak.com 2022-02-08 14:07 EDT-------
(In reply to comment #4)
> Please rebase eigen3 to 3.4 in 9.0 GA (if possible).
>
> Request: bz1870032 Comment 8
>
> The current version in RHEL 9 is 3.3.9. It will be good to do the rebase in
> 9.0 GA since it will be not possible doing it in minor releases (API
> changes, many new features, see release notes[1]).
>
> [1] Eigen 3.4 Release Notes: https://eigen.tuxfamily.org/index.php?title=3.4
>
> eigen3 is consumed by pybind11, which is consumed by scipy (CCing Matej
> Muzila and Nikola Forro).
>
>
> (In reply to Jiri Kucera from comment #2)
> > Tom, does your team has the capacity to handle the Eigen3 3.3.9 -> 3.4
> > rebase in 9.0 GA or it is too late?
>
> Hello Jirko,
> unfortunmately we do not plan accept more components in rhel9 and this one
> seems to be a leftover from our perspective. There is no other desktop pkg
> that would be dependen on this pkg. As you mentioned in c1 only pybind11. It
> would also make sense if we move Eigen3 under team which actually use it.

SO, I'm struggling here on the answers and priority..
On one hand RH indicates that we need to do rebase in 9.0, as it can't be done in rest of 9.x series due to API issues..
Is this correct?

From IBM side, if we're stuck at 3.3.9.. what's the impact?
Do we/industry care what's in RHEL vs delivered separately.?
What would this mean from back port load perspective both teams if changes needed?

------- Comment From Rajalakshmi.Srinivasaraghavan 2022-02-08 14:59 EDT-------
>> From IBM side, if we're stuck at 3.3.9.. what's the impact?

The POWER10 features are introduced only in eigen 3.4, especially optimization related to the Matrix Multiply Assist feature (MMA).

>> Do we/industry care what's in RHEL vs delivered separately.?

As eigen is a header-only package, this package can be easily cloned and used from the public repository, if they are not available in the RHEL repository. Since we have an existing package in RHEL with an older version, we raised this feature to update that.

>> What would this mean from back port load perspective both teams if changes needed?
There are a lot of changes and features on top of 3.3.9 and backporting would require more effort. As these are headers, it is not a suggested approach to backport, while these files can be easily replaced.

Hello, I forgot to mention here that update to eigen 3.4 is in progress, so if all will going smooth it will be shipped in RHEL 9.0 GA.

------- Comment From fnovak.com 2022-02-09 18:28 EDT-------
(In reply to comment #7)
> Hello, I forgot to mention here that update to eigen 3.4 is in progress, so
> if all will going smooth it will be shipped in RHEL 9.0 GA.

Great to hear..

Thanks for the help on this Jiri!

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory (new packages: eigen3), and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHBA-2022:4312