coupled models crash when MOM6 is compiled with -openmp about mom6

[Zhi-Liang]

Hi Niki,

I will take a look.

Zhi

On Tue, Jun 24, 2014 at 6:08 PM, Niki Zadeh [email protected]
wrote:

In branch dev/master/2014.06.23 (testing tag) all coupled test cases crash
when I compile with -openmp

Here's the traceback:

forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image PC Routine Line Source

fms_MOM6_SIS_LAD_ 000000000185D246 mom_tracer_hor_di 303
MOM_tracer_hor_diff.F90
fms_MOM6_SIS_LAD_ 0000000001717C50 mom_mp_step_mom_ 1062 MOM.F90
fms_MOM6_SIS_LAD_ 00000000016C2F27 ocean_model_mod_m 368
ocean_model_MOM.F90
fms_MOM6_SIS_LAD_ 00000000004013C7 MAIN__ 687 coupler_main.F90

-t repro does not have this problem.

To reproduce the issue:

fremake --execute --nolink -x
/ncrc/home2/Niki.Zadeh/frerts/testing_20140623/MOM6_CM.xml.20140624170350
-p ncrc2.intel -t repro-openmp MOM6_CM_libs_compile

fremake --execute -x
/ncrc/home2/Niki.Zadeh/frerts/testing_20140623/MOM6_CM.xml.20140624170350
-p ncrc2.intel -t repro-openmp MOM6_SIS_LAD_FV_compile

frerun -s -r basic -u --no-transfer -x
/ncrc/home2/Niki.Zadeh/frerts/testing_20140623/MOM6_CM.xml.20140624170350
-p ncrc2.intel -t repro-openmp MOM6_AM2_MOM6i_1deg

—
Reply to this email directly or view it on GitHub
#28.

[nikizadehgfdl]

These crashes always precede by a warning:

WARNING from PE    30: tracer_hordiff: Tracer T has associated 3-d diffusive flux diagnostics.  These are not valid when DIFFUSE_ML_TO_INTERIOR is defined. Use 2-d tracer diffusion diagnostics instead to get accurate total fluxes.

What does this mean?

[Zhi-Liang]

Hi Niki,

The non-openmp run also print out same warning message.

Zhi

On Wed, Jun 25, 2014 at 3:49 PM, Niki Zadeh [email protected]
wrote:

These crashes always precede by a warning:

WARNING from PE 30: tracer_hordiff: Tracer T has associated 3-d diffusive flux diagnostics. These are not valid when DIFFUSE_ML_TO_INTERIOR is defined. Use 2-d tracer diffusion diagnostics instead to get accurate total fluxes.

What does this mean?

—

Reply to this email directly or view it on GitHub
#28 (comment)
.

[Hallberg-NOAA]

Hi,
    This warning means that there are some diagnostic fields that
  are requested from the diag_table that simply can not be
  calculated with this particular model configuration. It is a
  warning, and it should not presage a model failure.
    - Bob Hallberg Dr. Robert Hallberg

Oceanographer and Ocean Group Head
NOAA Geophysical Fluid Dynamics Laboratory

NOAA GFDL || Phone: [phone]
Princeton University Forrestal Campus || Home: [phone]
201 Forrestal Road || Fax: [phone]
Princeton, New Jersey 08540-6649 || Email: [email protected]

  On 06/25/2014 03:46 PM, Niki Zadeh wrote:

  These crashes always precede by a warning:
    '''
    WARNING from PE 30: tracer_hordiff: Tracer T has associated 3-d
    diffusive flux diagnostics. These are not valid when
    DIFFUSE_ML_TO_INTERIOR is defined. Use 2-d tracer diffusion
    diagnostics instead to get accurate total fluxes.
    '''
    What does this mean?
  —
    Reply to this email directly or view
      it on GitHub.

[StephenGriffies]

"Presage" is a great word. We should all make a point of using it more
often.

On Wed, Jun 25, 2014 at 3:59 PM, Robert Hallberg [email protected]
wrote:

Hi,
This warning means that there are some diagnostic fields that
are requested from the diag_table that simply can not be
calculated with this particular model configuration. It is a
warning, and it should not presage a model failure.

• Bob Hallberg Dr. Robert Hallberg
  Oceanographer and Ocean Group Head
  NOAA Geophysical Fluid Dynamics Laboratory

NOAA GFDL || Phone: [phone]
Princeton University Forrestal Campus || Home: [phone]
201 Forrestal Road || Fax: [phone]
Princeton, New Jersey 08540-6649 || Email: [email protected]

On 06/25/2014 03:46 PM, Niki Zadeh wrote:

These crashes always precede by a warning:
'''
WARNING from PE 30: tracer_hordiff: Tracer T has associated 3-d
diffusive flux diagnostics. These are not valid when
DIFFUSE_ML_TO_INTERIOR is defined. Use 2-d tracer diffusion
diagnostics instead to get accurate total fluxes.
'''
What does this mean?
—
Reply to this email directly or view
it on GitHub.

—
Reply to this email directly or view it on GitHub
#28 (comment)
.

Dr. Stephen M. Griffies
NOAA Geophysical Fluid Dynamics Lab
201 Forrestal Road
Princeton, NJ 08542
USA

[Zhi-Liang]

Hi Niki,

I made a fix in MOM_tracer_hor_diff.F90. I change is to change argument
Reg from pointer to intent(inout).
I tested in /MOM6_AM2_MOM6i_1deg and it runs OK with openmp and reproduce
answer from no openmp.

Zhi

On Tue, Jun 24, 2014 at 6:08 PM, Niki Zadeh [email protected]
wrote:

In branch dev/master/2014.06.23 (testing tag) all coupled test cases crash
when I compile with -openmp

Here's the traceback:

forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image PC Routine Line Source

fms_MOM6_SIS_LAD_ 000000000185D246 mom_tracer_hor_di 303
MOM_tracer_hor_diff.F90
fms_MOM6_SIS_LAD_ 0000000001717C50 mom_mp_step_mom_ 1062 MOM.F90
fms_MOM6_SIS_LAD_ 00000000016C2F27 ocean_model_mod_m 368
ocean_model_MOM.F90
fms_MOM6_SIS_LAD_ 00000000004013C7 MAIN__ 687 coupler_main.F90

-t repro does not have this problem.

To reproduce the issue:

fremake --execute --nolink -x
/ncrc/home2/Niki.Zadeh/frerts/testing_20140623/MOM6_CM.xml.20140624170350
-p ncrc2.intel -t repro-openmp MOM6_CM_libs_compile

fremake --execute -x
/ncrc/home2/Niki.Zadeh/frerts/testing_20140623/MOM6_CM.xml.20140624170350
-p ncrc2.intel -t repro-openmp MOM6_SIS_LAD_FV_compile

frerun -s -r basic -u --no-transfer -x
/ncrc/home2/Niki.Zadeh/frerts/testing_20140623/MOM6_CM.xml.20140624170350
-p ncrc2.intel -t repro-openmp MOM6_AM2_MOM6i_1deg

—
Reply to this email directly or view it on GitHub
#28.

[adcroft]

What's the status on this issue?

[nikizadehgfdl]

It is fixed. Closing.

[Zhi-Liang]

Hi Alistair,

I think this issue is solved.

Zhi

On Tue, Aug 26, 2014 at 3:32 PM, Alistair Adcroft (GFDL) <
[email protected]> wrote:

What's the status on this issue?

—
Reply to this email directly or view it on GitHub
#28 (comment)
.