[Ncep.list.nems.announce] Proposed commit to NEMS trunk

John Michalakes john.michalakes at noaa.gov
Thu Feb 27 22:24:34 UTC 2014


These have been committed to the trunk as r37458 with Trac tickets #79 and
#80 now closed.  All clear...  Thanks, -John 
 
From: John Michalakes [mailto:john.michalakes at noaa.gov] 
Sent: Wednesday, February 26, 2014 4:32 PM
To: 'NEMS developers'
Subject: Proposed commit to NEMS trunk
 
Gentle NEMS developers,
 
This is my first time proposing an update to the NEMS trunk so please bear
with me.  I would like to commit the following two updates:
 
1.       Fix for an OpenMP thread safety issue
 
The Intel Inspector tool identifies a race condition for the DO J=JTS,JTE
loop at line 270 in atmos/phys/module_MP_FER_HIRES.F90 when running with
OMP_NUM_THREADS > 1.  The fix is to add the names of several more arrays to
the existing private declaration (see attached differences file).
 
2.       Parallel build for NEMS/NMMB
 
This set of proposed changes allows make to run in parallel when compiling
NEMS-NMMB (that is, when "make nmm" is invoked).  Haven't looked at the
other builds.  The affected makefiles are:
 
atmos/makefile
atmos/nmm/makefile
atmos/phys/makefile
atmos/share/makefile
makefile
 
If one does nothing, then "make nmm" builds the code serially in a litte
more than 11 minutes on WCOSS.  If one sets the environment variable:
 

     setenv J "-j8"   (csh/tcsh)
       -or-
     export J="-j8"   (bash)
 
the build will use as many as eight processes and compile in 4.5 minutes
(some of the dependencies force it to build serially or with only a couple
processes.  The phys directory is where it really flies.).  The integer that
follows the j (no space) is the number of processes that the make utility
should use.  If no integer is specified it will grab as many as it can (not
good).  Obviously, one wishes to avoid antagonizing other users on the
system (or worse, sysadmins) so one should avoid "going for broke" with this
setting.
 
The regression test is running now - sitting in the queue for TEST 009 as I
type this.  Assuming the test is a success, please review and let me know if
you have comments or questions.   Since this is my first time doing this,
I'm not sure what a "thumbs up" from the group looks like, so please let me
know that too.  
 
Complete set of differences with respect to the top-of-trunk is attached
here.  
 
Thanks,
 
John 
 
 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lstsrv.ncep.noaa.gov/pipermail/ncep.list.nems.announce/attachments/20140227/2554fc1a/attachment-0001.html 


More information about the Ncep.list.nems.announce mailing list