bugfix Mp38 allocatable array dimension by mefrediani · Pull Request #1834 · wrf-model/WRF

mefrediani · 2023-03-24T00:15:52Z

Large memory requirement in Thompson microphysics mp=8 and mp=28

TYPE: bug fix

KEYWORDS: memory, thompson, microphysics

SOURCE: Maria Frediani (NCAR)

DESCRIPTION OF CHANGES:
The new graupel (qg) table created for mp=38 required a larger memory to either compute or use. Because of that, when using options 8 or 28, the model stopped right away if there wasn't enough memory.

These changes make the 5D arrays truly allocatable, meaning when Thompson option 8 or 28 are used, the middle dimension (NRHG) is 1, and when 38 is used that dimension becomes 9. This should reduce the memory required when option 8 and 28 are used, and the newer qg table written out by the new code won't be as huge as it is now, but more similar to what it was before option 38 was introduced.

Also, the table was initially named qr_acr_qg_mp28V4, but this table is also used for option 8. For this reason, the name was changed to qr_acr_qg_V4.

LIST OF MODIFIED FILES: phys/module_mp_thompson.F

TESTS CONDUCTED:

Bit for bit comparison using diffwrf to compare the results between the current develop branch and the new code using mp=38 and mp=28
It passes the regression tests.

…anding height; option to print ember properties in rsl files; release location based on instant ROS

…ec. Change from 600 to 200

Trigger Jenkins tests

trigger Jenkins test

Remove lines to trigger a reg test.

weiwangncar · 2023-03-24T03:38:30Z

        if (.not. is_hail_aware) then
 !. Distribute the same value across NRHG dimension
-         do n3 = 2, NRHG
+         do n3 = 2, dimNRHG


@mefrediani If dimNRHG = 1, would it work? Maybe you could check the value as a condition to do this loop?

@mefrediani On second thought, perhaps this loop is no longer needed. What do you think?

Good catch, Wei! I can't believe I missed this!

Yes, probably no longer needed. Looks good so far. Hopefully this will solve the problem.

weiwangncar · 2023-03-24T15:47:28Z

@mefrediani Maybe we could take this opportunity to change the new table name for option 8 and 28 to be qr_acr_qg_V4.dat or something that does not imply option 28?

mefrediani · 2023-03-24T15:49:26Z

@weiwangncar I had already changed the table name ;)

weiwangncar · 2023-03-24T15:59:36Z

@mefrediani @dudhia I just tested with our obs-nudging case which used to stop right a way when 30 processors were used with one 36 processor node requested, it is now working! And the table size for option 8/28 has been reduced from 809,637,600 to 74,966,480 k-bytes! For reference, the qr_acr_qg table from prior WRF version is 89,959,776.

mefrediani · 2023-03-24T16:02:47Z

That's great, @weiwangncar! Thanks for finding a simple solution to the issue. Your directions were very helpful!

weiwangncar · 2023-03-24T16:05:01Z

@mefrediani You're welcome. I'm glad it is this simple. Do you know if the content in the new table for 8 and 28 has changed?

dudhia · 2023-03-24T16:17:02Z

Sounds good. Loop could be removed completely instead of left commented out.

…

On Fri, Mar 24, 2023 at 10:05 AM weiwangncar ***@***.***> wrote: @mefrediani <https://github.com/mefrediani> You're welcome. I'm glad it is this simple. Do you know if the content in the new table for 8 and 28 has changed? — Reply to this email directly, view it on GitHub <#1834 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEIZ77DD5GZGLMPDOZK5K2TW5XA3RANCNFSM6AAAAAAWF4LINA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

weiwangncar · 2023-03-24T16:42:33Z

The regression test results:

Test Type              | Expected  | Received |  Failed
= = = = = = = = = = = = = = = = = = = = = = = =  = = = =
Number of Tests        : 23           24
Number of Builds       : 60           57
Number of Simulations  : 158           150        0
Number of Comparisons  : 95           86        0

Failed Simulations are: 
None
Which comparisons are not bit-for-bit: 
None

mefrediani · 2023-03-24T17:45:11Z

@weiwangncar effectively, the tables in WRFv4.4 should provide the same values for mp28 and mp8. However, with this PR, we now have removed the variable tmg_gacr, so the tables will be different (that's why the new table is smaller).

weiwangncar · 2023-03-24T17:48:43Z

@mefrediani Thanks! I did realize the removal of the unused variable in the table. But it is good to know the content has not changed for options 8 and 28.

dudhia

loop is still there as commented out, but we can leave it

Remove a commented out loop. If needed, one can always find it in an older version of the code.

smileMchen · 2023-03-30T17:45:00Z

@mefrediani
Just rerun a nested case that failed previously. The case can run successfully with this fix. Many thanks !

mefrediani · 2023-03-30T18:38:37Z

@smileMchen I'm glad it worked! Thanks to @weiwangncar and the WRF team who suggested a simple solution to the issue.

Large memory requirement in Thompson microphysics mp=8 and mp=28 TYPE: bug fix KEYWORDS: memory, Thompson, microphysics SOURCE: Maria Frediani (NCAR) DESCRIPTION OF CHANGES: The new graupel (qg) table created for mp=38 required a larger memory to either compute or use (PR#1808). Because of that, when using options 8 or 28, the model stopped right away if there wasn't enough memory. These changes make the 5D arrays truly allocatable, meaning when Thompson option 8 or 28 are used, the middle dimension (NRHG) is 1, and when 38 is used that dimension becomes 9. This should reduce the memory required when option 8 and 28 are used, and the newer qg table written out by the new code won't be as huge as it is now, but more similar to what it was before option 38 was introduced. Also, the table was initially named qr_acr_qg_mp28V4, but this table is also used for option 8. For this reason, the name was changed to qr_acr_qg_V4. LIST OF MODIFIED FILES: phys/module_mp_thompson.F TESTS CONDUCTED: 1. Bit for bit comparison using diffwrf to compare the results between the current develop branch and the new code using mp=38 and mp=28. 2. It passes the regression tests.

mefrediani and others added 10 commits December 21, 2022 16:31

grass threat=1; fs_p_src stores ember tile source; option to change l…

b28af79

…anding height; option to print ember properties in rsl files; release location based on instant ROS

update start and solve in dyn_em

7e3ce13

decrease verbose - set wrfdbg=10

e4077a8

fix default: max_life_dt is given in number of timesteps, not total s…

3571b9f

…ec. Change from 600 to 200

fix write statement - convert it to INT

b61fa61

Update module_firebrand_spotting.F

1d88d0f

Trigger Jenkins tests

Update module_firebrand_spotting.F

df553a2

trigger Jenkins test

Update module_firebrand_spotting_mpi.F

799e052

Remove lines to trigger a reg test.

Merge branch 'wrf-model:develop' into develop

e71a126

NRHG allocatable dim

0a41bce

mefrediani requested a review from a team as a code owner March 24, 2023 00:15

weiwangncar reviewed Mar 24, 2023

View reviewed changes

fix do n3 = 2, dimNRHG

a36a3be

weiwangncar changed the base branch from develop to release-v4.5 March 24, 2023 16:01

dudhia previously approved these changes Mar 29, 2023

View reviewed changes

dudhia added the release-v4.5 label Mar 29, 2023

Remove commented out loop

812d04f

Remove a commented out loop. If needed, one can always find it in an older version of the code.

weiwangncar dismissed dudhia’s stale review via 812d04f March 30, 2023 17:34

weiwangncar approved these changes Mar 30, 2023

View reviewed changes

dudhia approved these changes Mar 31, 2023

View reviewed changes

weiwangncar merged commit 7825cd2 into wrf-model:release-v4.5 Mar 31, 2023

csmill mentioned this pull request Feb 11, 2026

Array out of bounds error in Thompson-aerosol aware scheme #2283

Open

Conversation

mefrediani commented Mar 24, 2023 • edited by weiwangncar Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

weiwangncar Mar 24, 2023

Choose a reason for hiding this comment

Uh oh!

weiwangncar Mar 24, 2023

Choose a reason for hiding this comment

Uh oh!

mefrediani Mar 24, 2023

Choose a reason for hiding this comment

Uh oh!

dudhia Mar 24, 2023

Choose a reason for hiding this comment

Uh oh!

weiwangncar commented Mar 24, 2023

Uh oh!

mefrediani commented Mar 24, 2023

Uh oh!

weiwangncar commented Mar 24, 2023

Uh oh!

mefrediani commented Mar 24, 2023

Uh oh!

weiwangncar commented Mar 24, 2023

Uh oh!

dudhia commented Mar 24, 2023 via email

Uh oh!

weiwangncar commented Mar 24, 2023

Uh oh!

mefrediani commented Mar 24, 2023

Uh oh!

weiwangncar commented Mar 24, 2023

Uh oh!

dudhia left a comment

Choose a reason for hiding this comment

Uh oh!

smileMchen commented Mar 30, 2023

Uh oh!

mefrediani commented Mar 30, 2023

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

mefrediani commented Mar 24, 2023 •

edited by weiwangncar

Loading