User/z1l/openmp #60

Zhi-Liang · 2014-09-24T17:42:29Z

Implement Openmp by using block structure.

The following shared codes update are needed,
cvs up -r ulm_block_z1l shared/mpp/include/mpp_domains_define.inc
cvs up -r ulm_block_z1l shared/mpp/mpp_domains.F90

The following files are changed.
src/core/MOM_PressureForce_Montgomery.F90
src/core/MOM_PressureForce_analytic_FV.F90
src/core/MOM_barotropic.F90
src/core/MOM_grid.F90
src/diagnostics/MOM_diagnostics.F90
src/equation_of_state/MOM_EOS.F90
src/equation_of_state/MOM_EOS_Wright.F90
src/equation_of_state/MOM_EOS_linear.F90
src/framework/MOM_domains.F90
src/framework/MOM_memory_macros.h

Zhi

adcroft · 2014-09-26T14:46:20Z

Zhi,

Evidently I need a patch to mpp to make this branch work. I thought everything would be in tikal_201409?

MOM6/src/framework/MOM_domains.F90:43.27:

use mpp_domains_mod, only : compute_block_extent => mpp_compute_block_extent 
                           1
Error: Symbol 'mpp_compute_block_extent' referenced at (1) not found in module 'mpp_domains_mod'
../../../../MOM6/src/framework/MOM_domains.F90:63.30:

public :: compute_block_extent
                              1
Error: Symbol 'compute_block_extent' at (1) has no IMPLICIT type
make[3]: *** [MOM_domains.o] Error 1
make[3]: *** Waiting for unfinished jobs....
make[3]: Leaving directory `/lustre/f1/dev/gfdl_O/Alistair.Adcroft/MOM6.workdir/build/gnu/solo_ocean/repro'
make[2]: *** [build/gnu/solo_ocean/repro/MOM6] Error 2
make[2]: Leaving directory `/lustre/f1/dev/gfdl_O/Alistair.Adcroft/MOM6.workdir'

Zhi-Liang · 2014-09-26T16:14:48Z

Hi Alistair,

Yes, it need the following shared code update.
cvs up -r ulm_block_z1l shared/mpp/include/mpp_domains_define.inc
cvs up -r ulm_block_z1l shared/mpp/mpp_domains.F90

Thanks,

Zhi

On Fri, Sep 26, 2014 at 10:46 AM, Alistair Adcroft (GFDL) <
[email protected]> wrote:

Zhi,

Evidently I need a patch to mpp to make this branch work. I thought
everything would be in tikal_201409?

MOM6/src/framework/MOM_domains.F90:43.27:

use mpp_domains_mod, only : compute_block_extent => mpp_compute_block_extent
1
Error: Symbol 'mpp_compute_block_extent' referenced at (1) not found in module 'mpp_domains_mod'
../../../../MOM6/src/framework/MOM_domains.F90:63.30:

public :: compute_block_extent
1
Error: Symbol 'compute_block_extent' at (1) has no IMPLICIT type
make[3]: *** [MOM_domains.o] Error 1
make[3]: *** Waiting for unfinished jobs....
make[3]: Leaving directory /lustre/f1/dev/gfdl_O/Alistair.Adcroft/MOM6.workdir/build/gnu/solo_ocean/repro' make[2]: *** [build/gnu/solo_ocean/repro/MOM6] Error 2 make[2]: Leaving directory/lustre/f1/dev/gfdl_O/Alistair.Adcroft/MOM6.workdir'

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

adcroft · 2014-09-26T18:26:20Z

I don't see this at https://github.com/CommerceGov/NOAA-GFDL-FMS. We need the patch to be available for external users before I can add it to dev/master otherwise they'll be stuck. Can you add it there?

underwoo · 2014-09-26T20:39:15Z

Currently I am the only one that can make modifications to github FMS. I
will make the modification when I return on Monday.

On Friday, September 26, 2014, Alistair Adcroft (GFDL) <
[email protected]> wrote:

I don't see this at https://github.com/CommerceGov/NOAA-GFDL-FMS. We need
the patch to be available for external users before I can add it to
dev/master otherwise they'll be stuck. Can you add it there?

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

Seth Underwood
Engility
High Technology Services Group

Modeling Services Group
GFDL/NOAA/DOC
201 Forrestal Road
Princeton, NJ 08540-6649

(609) 452-5847 Office
(304) 376-9002 Cell
(609) 987-5063 Fax
[email protected]

underwoo · 2014-09-29T15:22:06Z

I have put the CVS tag ulm_block_z1l under the git branch
user/z1l/ulm_block.

Seth Underwood
Engility
High Technology Services Group

Modeling Services Group
GFDL/NOAA/DOC
201 Forrestal Road
Princeton, NJ 08540-6649

(609) 452-5847 Office
(304) 376-9002 Cell
(609) 987-5063 Fax
[email protected]

On Fri, Sep 26, 2014 at 4:39 PM, Seth Underwood - NOAA Affiliate <
[email protected]> wrote:

Currently I am the only one that can make modifications to github FMS. I
will make the modification when I return on Monday.

On Friday, September 26, 2014, Alistair Adcroft (GFDL) <
[email protected]> wrote:

I don't see this at https://github.com/CommerceGov/NOAA-GFDL-FMS. We
need the patch to be available for external users before I can add it to
dev/master otherwise they'll be stuck. Can you add it there?

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

Seth Underwood
Engility
High Technology Services Group

Modeling Services Group
GFDL/NOAA/DOC
201 Forrestal Road
Princeton, NJ 08540-6649

(609) 452-5847 Office
(304) 376-9002 Cell
(609) 987-5063 Fax
[email protected]

adcroft · 2014-10-09T14:25:04Z

I just tried this branch again and have several failures when not using openmp:

Different solution in examples/solo_ocean/mixed_layer_restrat_2d
SIGSEGV with gnu in examples/ocean_SIS/OM4_025
(also, using intel the model seems to hang somewhere, didn't try pgi)
SIGSEGV with gnu in examples/ocean_SIS2/OM4_025

Zhi-Liang · 2014-10-09T14:44:55Z

Hi Alistair,

I will take a look.

Zhi

On Thu, Oct 9, 2014 at 10:25 AM, Alistair Adcroft (GFDL) <
[email protected]> wrote:

I just tried this branch again and have several failures when not using
openmp:

Different solution in examples/solo_ocean/mixed_layer_restrat_2d

SIGSEGV with gnu in examples/ocean_SIS/OM4_025 (also, using intel
the model seems to hang somewhere, didn't try pgi)

SIGSEGV with gnu in examples/ocean_SIS2/OM4_025

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

Zhi-Liang · 2014-10-09T18:04:36Z

Hi Alistair,

There is a careless index bug in MOM_memory_macros.h and I just fixed it
and pushed the fix in github. I tested all the three tests and all could
reproduce dev/master ( For ocean_SIS2/OM4_025, timestats could not
reproduce the one repository but it reproduce the answer I compile with
dev/master code).

Greetings,

Zhi

On Thu, Oct 9, 2014 at 10:44 AM, Zhi Liang - NOAA Affiliate <
[email protected]> wrote:

Hi Alistair,

I will take a look.

Zhi

On Thu, Oct 9, 2014 at 10:25 AM, Alistair Adcroft (GFDL) <
[email protected]> wrote:

I just tried this branch again and have several failures when not using
openmp:

Different solution in examples/solo_ocean/mixed_layer_restrat_2d

SIGSEGV with gnu in examples/ocean_SIS/OM4_025 (also, using intel
the model seems to hang somewhere, didn't try pgi)

SIGSEGV with gnu in examples/ocean_SIS2/OM4_025

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

Zhi-Liang · 2014-10-09T18:07:24Z

Hi Alistair,

This bug only cause problem when number of points in y-direction
computation domain is greater than number of points in x-direction. That is
the reason that I ran 20+ solo tests and all passed the test.

Zhi

On Thu, Oct 9, 2014 at 2:04 PM, Zhi Liang - NOAA Affiliate <
[email protected]> wrote:

Hi Alistair,

There is a careless index bug in MOM_memory_macros.h and I just fixed it
and pushed the fix in github. I tested all the three tests and all could
reproduce dev/master ( For ocean_SIS2/OM4_025, timestats could not
reproduce the one repository but it reproduce the answer I compile with
dev/master code).

Greetings,

Zhi

On Thu, Oct 9, 2014 at 10:44 AM, Zhi Liang - NOAA Affiliate <
[email protected]> wrote:

Hi Alistair,

I will take a look.

Zhi

On Thu, Oct 9, 2014 at 10:25 AM, Alistair Adcroft (GFDL) <
[email protected]> wrote:

I just tried this branch again and have several failures when not using
openmp:

Different solution in examples/solo_ocean/mixed_layer_restrat_2d

SIGSEGV with gnu in examples/ocean_SIS/OM4_025 (also, using intel
the model seems to hang somewhere, didn't try pgi)

SIGSEGV with gnu in examples/ocean_SIS2/OM4_025

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

…_cvmix Double diffusion cvmix + additional updates

Zhi-Liang added 4 commits September 12, 2014 16:51

Fix the potential crash when nstep+nfilter=0

c2530ba

Merge remote-tracking branch 'origin' into user/z1l/openmp

5acdd71

implement openmp using block structure

b1d7687

Merge remote-tracking branch 'origin' into user/z1l/openmp

9edf4bf

Zhi-Liang added 2 commits October 9, 2014 11:23

Merge remote-tracking branch 'origin' into user/z1l/openmp

af77e3f

Bug fix

4d1b54c

adcroft merged commit 4d1b54c into dev/master Oct 9, 2014

gustavo-marques added a commit to gustavo-marques/MOM6 that referenced this pull request May 23, 2018

Merge pull request mom-ocean#60 from gustavo-marques/double_diffusion…

ec95be9

…_cvmix Double diffusion cvmix + additional updates

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

User/z1l/openmp #60

User/z1l/openmp #60

Zhi-Liang commented Sep 24, 2014

adcroft commented Sep 26, 2014

Zhi-Liang commented Sep 26, 2014

adcroft commented Sep 26, 2014

underwoo commented Sep 26, 2014

underwoo commented Sep 29, 2014

adcroft commented Oct 9, 2014

Zhi-Liang commented Oct 9, 2014

Zhi-Liang commented Oct 9, 2014

Zhi-Liang commented Oct 9, 2014

User/z1l/openmp #60

User/z1l/openmp #60

Conversation

Zhi-Liang commented Sep 24, 2014

adcroft commented Sep 26, 2014

Zhi-Liang commented Sep 26, 2014

adcroft commented Sep 26, 2014

underwoo commented Sep 26, 2014

underwoo commented Sep 29, 2014

adcroft commented Oct 9, 2014

Zhi-Liang commented Oct 9, 2014

Zhi-Liang commented Oct 9, 2014

Zhi-Liang commented Oct 9, 2014