Develop including FB_pdlib_changes from Mathieu

[thesser1]

Reg tests have been completed. I have attached the resulting files.
FB_pdlib_changes.zip

"This PR yields smaller memory usage for the PDLIB and is an unambiguous win.
Let me know if there are any problem.
Mathieu"

[ajhenrique]

@JessicaMeixner-NOAA and @mickaelaccensi please review this pull request. Thanks for your help!

[mickaelaccensi]

sounds ok for me since the regtests are correctly passed

[JessicaMeixner-NOAA]

Based on comparing with https://github.com/NOAA-EMC/WW3/wiki/How-to-use-matrix.comp-to-compare-regtests-with-master#4-look-at-results
The tests mww3_test_04/./work_PR3_UQ_MPI_NC_b_c has files that are different but not on the list of expected differences (likely this is a new test and the netcdf files are not being compared properly by the script from what it maybe looks like?) Also mww3_test_03/./work_PR2_UQ_MPI_d2 test has double the expected number of different files. There were lots of tests with the a file gmon.out that is different? What is this file?

The changes themselves are in PDLIB which I don't know well enough to judge if the code is good/bad. The PDLIB regression tests passed with the exception of the gmon.out file, so depending on what that file is I'm assuming everything is okay.

[mickaelaccensi]

Hi Jessica,
gmon.out is a profiling file which can be usefull to track where the run took the most part of its time. This file should be excluded from the regtest comparison.

[JessicaMeixner-NOAA]

I created an issue to remove the gmon.out file. Again, don't think I'm the most qualified to review this but from what I can tell it's okay. Aron would be a better reviewer. Or Andre or Ali who use PDLIB more?

[mickaelaccensi]

It's not part of PDLIB, it's based on the compilation options for profiling "-p"

[thesser1]

Thanks. I will look into it. Ty

…

On Fri, Aug 23, 2019 at 2:22 AM Mickael Accensi ***@***.***> wrote: It's not part of PDLIB, it's based on the compilation options for profiling "-p" — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#84?email_source=notifications&email_token=AAU2O3BA3MNLWQX4E7C3BJDQF562XA5CNFSM4INA3VV2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD47HX2Q#issuecomment-524188650>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAU2O3CN6VNU4VXEHXUSISLQF562XANCNFSM4INA3VVQ> .

[JessicaMeixner-NOAA]

Yes - The gmon.out is not pdlib related, but the updates to the actual code are pdlib related.

[ajhenrique]

@thesser1 @aliabdolali please address Issue #87 (add files gmon and time_count.txt in files to be excluded from matrix_comp comparisons in parameter skipfiles L162).

[ajhenrique]

@thesser1 @aliabdolali have you made progress on ironing out the pending issues to get this PR merged to our develop branch? Thanks.

[ajhenrique]

@JessicaMeixner-NOAA has provided a fix for issue #87 at
6c8c0f0

[thesser1]

I reran the matrix with the rollback develop on the erdc repo since I had already merge the FB_pdlib branch in the erdc develop branch. The test returned some new differences. I am now running the matrix on the EMC develop branch to make sure I understand what is going on.
FB_pdlib_changes_v2.zip

[aliabdolali]

@thesser1 @ajhenrique
Hi Henrique,
I checked the report, looks good to me, I can merge it or you do it.
Regards
AA

[ajhenrique]

@thesser1 @ajhenrique
Hi Henrique,
I checked the report, looks good to me, I can merge it or you do it.
Regards
AA

Hi @aliabdolali @thesser1 I was not able to open the report, I would like to have a look at it as well. @thesser1 can you explain what new differences were there, @aliabdolali why do you think they are not relevant. Thanks!

[ajhenrique]

@thesser1 @aliabdolali @mickaelaccensi @JessicaMeixner-NOAA I ran the regtest matrix on Hera, our new R&D machine for both an update for porting the package to the new resource and for the ERDC PR. The differences reporduced those pointed our by Jessica in the thread above. I will assume all is OK, however need that @mickaelaccensi approves this PR as a reviewer to complete the process, then I'll merge this PR to the NOAA-EMC/WW3 develop.

[ajhenrique]

Completed regression testing matrix using NCEP's R&D resource Hera (Intel SkyLake 2.40GHz cpus, on 1 node/40cpus). Results confirm those reported by @thesser1 and @aliabdolali and insights provided by @JessicaMeixner-NOAA. The following compiler and modules were used: intel/14.0.2 impi/5.1.2.150 netcdf/4.3.0.