Fix model slowness when using threading

[climbfuji]

Description

This PR fixes the model slowness that we started to see after PR #249 was merged. The reason are missing OpenMP dependencies in ccpp/data/CMakeLists.txt and ccpp/driver/CMakeLists.txt.

These dependencies were missing since the two folders ccpp/data and ccpp/driver were created. However, OpenMP was added implicitly through a dependency on target gfsphysics, which was removed in PR #249.

Issue(s) addressed

ufs-weather-model issue #474

Testing

Tested manually with Intel and GNU on Cheyenne and Hera for the regression test fv3_ccpp_2threads for 240h long integration times. With these changes the performance of the pre-#249 code is regained (and at the same time the memory footprint is reduced due to the changes in #249).

Dependencies

n/a

[aerorahul]

were we not just still having a discussion on the PUBLIC and PRIVATE nature of OpenMP linking in another thread?

[climbfuji]

Yes, but this can be done in my next commit (tomorrow) or later if we can't come to a conclusion by then. I will follow your advise as to what we should use, I don't understand cmake as well as you do.

…

On Mar 16, 2021, at 7:37 AM, Rahul Mahajan ***@***.***> wrote: were we not just still having a discussion on the PUBLIC and PRIVATE nature of OpenMP linking in another thread? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#258 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB5C2RODSAFGRYM7B2NLWJDTD5NH3ANCNFSM4ZC73K4Q>.

[junwang-noaa]

I think the slowness issue is resolved by the code changes. The update on the PUBLIC and PRIVATE nature of OpenMP may require more changes besides the changes in this PR, e.g. FV3/CMakeLists.txt, On Tue, Mar 16, 2021 at 9:39 AM Dom Heinzeller ***@***.***> wrote:

…

Yes, but this can be done in my next commit (tomorrow) or later if we can't come to a conclusion by then. I will follow your advise as to what we should use, I don't understand cmake as well as you do. > On Mar 16, 2021, at 7:37 AM, Rahul Mahajan ***@***.***> wrote: > > > were we not just still having a discussion on the PUBLIC and PRIVATE nature of OpenMP linking in another thread? > > — > You are receiving this because you authored the thread. > Reply to this email directly, view it on GitHub < #258 (comment)>, or unsubscribe < https://github.com/notifications/unsubscribe-auth/AB5C2RODSAFGRYM7B2NLWJDTD5NH3ANCNFSM4ZC73K4Q >. > — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#258 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AI7D6TLHVEN3WER5VEI3JMDTD5NSJANCNFSM4ZC73K4Q> .