UKMO Staging jan2021 #327

ukmo-ccbunney · 2021-03-05T09:38:25Z

The staging branch comprises two PRs from the ukmo-waves repository:

Bug fix and small simplification/optimization change ukmo-waves/WW3#18 : Bug fix for ww3_multi hanging issue
First set of changes intended to fix the bug ukmo-waves/WW3#19 : Bug fix for interpolation weights calculation in ww3_prnc.

Fixes #290
Fixes #207
Fixes #314

Full regression test output for each PR are available in the PRs links above.

Regression test outputs for this combined staging branch will be provided below.

Fixes: NOAA-EMC#314 * Interpolation weights now correctly calculated on points next to land and BC locations. * Changes to improve the code: the possibility of reading zero values from the input is considered, and points that should not be taken into account in the interpolation are identified by the netcdf fill value; a subroutine is created to avoid code duplication

* Fixes NOAA-EMC#290 (ww3_multi hanging when generating restart with IOSTYP >= 2) * Also fixes out-of-bounds array access error. * Includes some MPI optimizations

JessicaMeixner-NOAA · 2021-03-05T13:24:06Z

model/ftn/w3iorsmd.ftn

@@ -609,7 +604,7 @@
 !/MPI                          IF ( IP .EQ. NAPRST ) THEN
 !/MPI                              WRITEBUFF(1:NSPEC) = VA(1:NSPEC,JSEA)
 !/MPI                            ELSE
-!/MPI                              JSEA   = JSEA - 2*((IB-1)/2)*RSBLKS
+!/MPI                              JSEA = 1 + MOD(IB-1,2)*RSBLKS


Do the bug fixes here warrant changing the date for this file?

I was going to say no, as it as it doesn't actually change the format of the file... but I suppose it could change how an earlier file was read in w.r.t. JSEA? @ukmo-juan-castillo - can you confirm?

JessicaMeixner-NOAA · 2021-03-05T13:26:34Z

Does this close any issues? I want to say there is at least one issue this PR would close.

to take into account the coupled configurations, that are also affected

ukmo-ccbunney · 2021-03-11T11:16:01Z

Does this close any issues? I want to say there is at least one issue this PR would close.

@JessicaMeixner-NOAA : yes - I've edited the description to link in the relevant issues.

in w3iorsmd

ukmo-ccbunney · 2021-03-17T12:27:11Z

@ukmo-juan-castillo and I are happy with the regtest results for this PR now. The output of the full regression test maxtrix compared against develop is here:

matrixComp_staging_jan2021.zip

Summary of differences relavent to this PR:

mww3_test_04: As expected only the ‘d’ test where we changed the test to write the restart change results
ww3_tp2.14: Differences in r-toy.nc and r-ww3.nc are caused OASIS, appending to files rather than overwriting them when a test is re-run; reproducibility is something I want to investigate later with Jessica
ww3_tp15: as expected due to corrections in the interpolation of netcdf input fields

Also, the usual culprits which have known differences

ww3_tp2.18: Existing known differences in output to screen
mww3_test_03: Known to be not b4b.

Over to you @aliabdolali :)

aliabdolali · 2021-03-17T14:58:05Z

Hi @ukmo-juan-castillo and @ukmo-ccbunney
I am going to check it at our end today.

aliabdolali · 2021-03-18T14:32:39Z

Hi @ukmo-juan-castillo and @ukmo-ccbunney
The output of the full regression test maxtrix compared against develop is here:
MatrixDiffNCEP.zip
I have similar differences as you mentioned. tp2.14 are identical on our HPC and the ww3_tp2.10 has always differences on our HPC. The rests are the same.

ukmo-ccbunney · 2021-03-18T15:47:50Z

Hi @ukmo-juan-castillo and @ukmo-ccbunney
The output of the full regression test maxtrix compared against develop is here:
MatrixDiffNCEP.zip
I have similar differences as you mentioned. tp2.14 are identical on our HPC and the ww3_tp2.10 has always differences on our HPC. The rests are the same.

@aliabdolali I notice that you have some differences in the mod_def.* files for mww3_test_07. I wasn't expecting to see any differences there (and we have not in our tests). Do you have any idea what is happening there (it must be some insignificant difference as all the other output files are identical for that test)?

Everything else looks as expected.

aliabdolali · 2021-03-18T17:55:30Z

Hi @ukmo-juan-castillo and @ukmo-ccbunney
The output of the full regression test maxtrix compared against develop is here:
MatrixDiffNCEP.zip
I have similar differences as you mentioned. tp2.14 are identical on our HPC and the ww3_tp2.10 has always differences on our HPC. The rests are the same.

@aliabdolali I notice that you have some differences in the mod_def.* files for mww3_test_07. I wasn't expecting to see any differences there (and we have not in our tests). Do you have any idea what is happening there (it must be some insignificant difference as all the other output files are identical for that test)?

Everything else looks as expected.

@ukmo-ccbunney I checked another PR and I had similar non-identical results in mod_def.* files for mww3_test_07. So it is intel/our HPC related.

ukmo-ccbunney · 2021-03-19T08:55:13Z

@ukmo-ccbunney I checked another PR and I had similar non-identical results in mod_def.* files for mww3_test_07. So it is intel/our HPC related.

Phew - that's a relief. Also - this is interesting behaviour as sometimes I also have differences in the mod_def.ww3 files for one particular regtest (I forget which one now). I think it might be a compiler issue as the differences usually go away if the filesystem path to the WW3 installation directory has the same number of characters (yes - bizarre I know!!); hence I think it is a GNU Fortran bug/oddity*

* I sometimes forget that compilers are written by humans too!

JessicaMeixner-NOAA · 2021-03-19T13:19:52Z

@ukmo-ccbunney I remember we had this happen before because there was an uninitialized variable being stored in the mod_def.ww3 file. I'm not sure if it's the same file/test or not, but something like that could explain why some compilers it's fine and others it is not.

ukmo-ccbunney · 2021-03-19T15:12:49Z

Fixes #206

ukmo-juan-castillo added 2 commits March 2, 2021 09:43

Bug fix and small simplification/optimization change (#18)

6930fac

* Fixes NOAA-EMC#290 (ww3_multi hanging when generating restart with IOSTYP >= 2) * Also fixes out-of-bounds array access error. * Includes some MPI optimizations

ukmo-ccbunney requested a review from aliabdolali March 5, 2021 09:38

ukmo-ccbunney added the bug Something isn't working label Mar 5, 2021

ukmo-ccbunney marked this pull request as draft March 5, 2021 12:42

JessicaMeixner-NOAA reviewed Mar 5, 2021

View reviewed changes

Correction to the bug fix in branch bf_multi_hang

e4f8f51

to take into account the coupled configurations, that are also affected

Small correction to the multi_hang branch: revert changes to JSEA index

c808c3e

in w3iorsmd

ukmo-ccbunney marked this pull request as ready for review March 17, 2021 12:19

aliabdolali requested a review from JessicaMeixner-NOAA March 19, 2021 13:15

aliabdolali approved these changes Mar 19, 2021

View reviewed changes

JessicaMeixner-NOAA approved these changes Mar 19, 2021

View reviewed changes

aliabdolali merged commit 6d95c33 into NOAA-EMC:develop Mar 19, 2021

ukmo-ccbunney deleted the staging_jan2021 branch May 25, 2021 13:51

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

UKMO Staging jan2021 #327

UKMO Staging jan2021 #327

ukmo-ccbunney commented Mar 5, 2021 •

edited

Loading

JessicaMeixner-NOAA Mar 5, 2021

ukmo-ccbunney Mar 5, 2021

JessicaMeixner-NOAA commented Mar 5, 2021

ukmo-ccbunney commented Mar 11, 2021

ukmo-ccbunney commented Mar 17, 2021

aliabdolali commented Mar 17, 2021

aliabdolali commented Mar 18, 2021

ukmo-ccbunney commented Mar 18, 2021 •

edited

Loading

aliabdolali commented Mar 18, 2021 •

edited

Loading

ukmo-ccbunney commented Mar 19, 2021 •

edited

Loading

JessicaMeixner-NOAA commented Mar 19, 2021

ukmo-ccbunney commented Mar 19, 2021

UKMO Staging jan2021 #327

UKMO Staging jan2021 #327

Conversation

ukmo-ccbunney commented Mar 5, 2021 • edited Loading

JessicaMeixner-NOAA Mar 5, 2021

Choose a reason for hiding this comment

ukmo-ccbunney Mar 5, 2021

Choose a reason for hiding this comment

JessicaMeixner-NOAA commented Mar 5, 2021

ukmo-ccbunney commented Mar 11, 2021

ukmo-ccbunney commented Mar 17, 2021

aliabdolali commented Mar 17, 2021

aliabdolali commented Mar 18, 2021

ukmo-ccbunney commented Mar 18, 2021 • edited Loading

aliabdolali commented Mar 18, 2021 • edited Loading

ukmo-ccbunney commented Mar 19, 2021 • edited Loading

JessicaMeixner-NOAA commented Mar 19, 2021

ukmo-ccbunney commented Mar 19, 2021

ukmo-ccbunney commented Mar 5, 2021 •

edited

Loading

ukmo-ccbunney commented Mar 18, 2021 •

edited

Loading

aliabdolali commented Mar 18, 2021 •

edited

Loading

ukmo-ccbunney commented Mar 19, 2021 •

edited

Loading