Routine naming conventions: rook codes

[zerothi]

The newly added rook3 codes which is based on BLAS-level 3 are named *_rk while the older rook routines are named *_rook.

As I suggested in #82, I think they should be named more similarly (here is a suggestion which may not at all reflect their intend and actual code):

1. rename _rook to _rook2, and rename _rk to rook3?
2. rename _rook to _rookkb2, and rename _rk to rookkb3? (rook-Bunch-Kaufman)
3. rename _rook to _rkkb2, and rename _rk to rkkb3? (rook-Bunch-Kaufman)
4. rename _rk to rook3

[julielangou]

We started the discussion with the team... Your request makes a lot of sense, yet the team did not make a decision yet. We will keep you posted.

[zerothi]

Great, thanks.
I think it is vital that there will be some kind of guideline for new algorithms. I have also noticed your Aasen routine changes, perhaps some-kind of generic template for generating the routine names will be welcoming for the future. In any case I wait in anticipation ;)

[langou]

Hi Nick. Sorry for not coming back to you. We did discuss this (thanks to your input), a recommendation was made and I never came back to you to report. First of, you need to know that the plan is to ultimately get rid of _rook. We might move _rook in DEPRECATED for now, but the goal is to remove _rook from LAPACK. Yet, since the interfaces of _rook and _rk are different, we did not want to use the _rook name for _rk. So we decided to go with _rk for _rook. (Well, _rk and _rook seemed equally good so this last point tipped the balance if I recall correctly.) We went for _aa for Aasen. Instead of a possible _aasen. Happy to discuss this some more but the current status is (1) _aa for Aasen, (2) _rk for new rook, and (3) deprecate _rook. I hope this makes sense. How do you feel about this? Thanks for the input in any case, Cheers, Julien.

[zerothi]

No worries!
Whatever you decide is fine by me :)
Your proposed scheme makes it consistent which is what I am concerned about ;)

PS. I will let this be open so that you may close it in a PR/commit.

[langou]

Hi Nick, I am going to close this thread. I think we are all good with the code as is. Thanks for the input. Cheers, Julien.

[zerothi]

Ok, thanks for taking it into consideration!

…

On 5 Dec 2016 9:27 pm, "langou" ***@***.***> wrote: Hi Nick, I am going to close this thread. I think we are all good with the code as is. Thanks for the input. Cheers, Julien. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#91 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AEkWIcXv1aUOHua7GVjfHBRkHyD44OE9ks5rFHPIgaJpZM4K5hJG> .