[Ncep.list.nems.announce] nems trunk commit

[Raghu Reddy]

Hi Dusan,

I'm not sure if it is going to be a major issue or not, but I do have one
concern and a suggestion about parallel makes that I would like to mention.

Most of the time that regression test script is typically run on the front
end, and multiple users running them simultaneously (or with significant
overlap) can put a significant load on the frontends and adversely affect a
number of users.

Like I mentioned in the beginning, this may or may not be a concern.  So a
couple of questions are:

What is the default value of J (the number of parallel compiles)?  
Is it possible for the compilation step also be submitted as a batch job to
a compute node?

Submitting the compilation step as a batch job has a couple of advantages:

One can use all the cores on that node to compile in parallel.  

And the second most significant benefit is the fact that on compute nodes
/tmp is a memory resident file system, and so compilations are likely to
progress much faster than on the frontends which has a physical disk for
/tmp.

These two things together should help significantly in reducing the
compilation time.

Since batch jobs so are already being used for executing the test cases,
hopefully it shouldn't be too much to also do the competition step in a
batch job?

Just a suggestion.

Thanks,
Raghu



-----Original Message-----
From: ncep.list.nems.announce-bounces at lstsrv.ncep.noaa.gov
[mailto:ncep.list.nems.announce-bounces at lstsrv.ncep.noaa.gov] On Behalf Of
Dusan Jovic
Sent: Thursday, October 02, 2014 8:54 AM
To: ncep.list.nems.announce at lstsrv.ncep.noaa.gov
Subject: [Ncep.list.nems.announce] nems trunk commit

Hello,

  I have a small set of changes that I'd like to commit to the NEMS trunk.
In NMMB the changes have been made to allow 3-digit level number in 3D
variable names which will allow model to run with >99 vertical levels. Then
there are number of small changes in regtests. Parallel compilation is used
now (added by John M. some time ago) in rt.sh.  
Fixed comments in esmf_configure. Fixed the problem that occasionally caused
nmmb tests to time-out on zeus. Enabled 'thompson microphysics' 
regresison test ( inadvertently commented out during one of the previous
commits).

Here is the svn status:


Dusan.Jovic at fe5:/scratch2/portfolios/NCEPDEV/meso/save/Dusan.Jovic/trunk_lev
3>
svn status .
M       tests/RegressionTests_zeus.log
M       tests/rt.sh
M       tests/rt_nmm.sh
M       tests/rt.conf
M       tests/Compile_zeus.log
M       tests/Compile_wcoss.log
M       tests/RegressionTests_wcoss.log
X       src/atmos/gsm
M       src/atmos/post/post_nmm.f
M       src/atmos/nmm/module_RADIATION.F90
M       src/atmos/nmm/module_WRITE_GRID_COMP.F90
M       src/atmos/nmm/module_WRITE_ROUTINES.F90
M       src/atmos/nmm/module_VARS_STATE.F90
M       src/conf/modules.nems.wcoss_ESMF_630rAPI_gsm
M       src/conf/modules.nems.wcoss_ESMF_630rAPI_nmm
M       src/conf/modules.nems.wcoss_ESMF_3
M       src/conf/modules.nems.Zeus_ESMF_310rAPI
M       src/conf/modules.nems.Zeus_ESMF_520rAPI
M       src/conf/modules.nems.Zeus_ESMF_630rAPI
M       src/conf/modules.nems.Zeus_ESMF_NUOPC
M       src/esmf_version

Performing status on external item at 'src/atmos/gsm'
Dusan.Jovic at fe5:/scratch2/portfolios/NCEPDEV/meso/save/Dusan.Jovic/trunk_lev
3> 



Code passed full regression test on both machines. My workimg copies are::

zeus:/scratch2/portfolios/NCEPDEV/meso/save/Dusan.Jovic/trunk_lev3
tide:/meso/save/Dusan.Jovic/trunk_lev3

Dusan
_______________________________________________
Ncep.list.nems.announce mailing list
Ncep.list.nems.announce at lstsrv.ncep.noaa.gov
https://lstsrv.ncep.noaa.gov/mailman/listinfo/ncep.list.nems.announce