Running the GEOS-5 SBU Benchmark
Method for Checking Out GEOS-5 GCM
Set up CVSROOT
First, set up your CVSROOT environment variable using the scheme provided NCCS's CVSACL webpage (requires NCCS login) where:
On Discover and NAS: CVSROOT=:ext:$USER@cvsacldirect:/cvsroot/esma Elsewhere: CVSROOT=:ext:$USER@ctunnel:/cvsroot/esma
Set up CVS_RSH
Also, set CVS_RSH=ssh in your environment.
Start the tunnel (machines other than Discover and Pleiades)
Checking out the model
Make the directory in which you wish to checkout the model:
$ mkdir Heracles-UNSTABLE $ cd Heracles-UNSTABLE/
And do the actual checkout using:
$ cvs co -r Heracles-UNSTABLE GEOSagcm
In general:
$ cvs co -r <Tag Name> <Module Name>
where <Tag Name> is the tag for the model to check out (e.g., Heracles-UNSTABLE, Ganymed-4_1) and <Module Name> is the module (e.g., GEOSagcm usually, though some older tags use: Ganymed, Fortuna).
Build and install the model
First, set up ESMADIR:
$ setenv ESMADIR <directory-to>/Heracles-UNSTABLE/GEOSagcm
it is just below the src/ directory.
Go into the src/ directory of your model. Following above:
$ cd $ESMADIR/src
Setup the environment by sourcing the g5_modules file:
$ source g5_modules
To build the model, you have one of two choices. First, you can use the parallel_build.csh script to submit a PBS job that compiles the model:
$ ./parallel_build.csh
or you can interactively build the model using:
$ gmake install
To capture the install log, we recommend tee'ing the output to a file:
$ gmake install |& tee make.install.log (on tcsh) $ gmake install 2>&1 | tee make.install.log (on bash)
Note you can also build in parallel interactively with:
$ gmake --jobs=N pinstall |& tee make.pinstall.log (on tcsh)
where N is the number of parallel processes. For best performance, N should never be, say, more than 2 less than the number of cores. So, on a Haswell node, you could use 26, but any number over about 10 is probably not worth it. Note: for the sake of others, do this on an interactive node if you are going to use 16 cores or so. If you are on a head node, a 2-4 core pinstall is probably safe and shouldn't bother too many people.
By default, the Intel Fortran compiler (ifort) is used for the build process. For other compilers, contact matthew.thompson@nasa.gov for instructions to use GCC or PGI compilers.
Monitor build process
The build can be monitored using the utility gmh.pl in the directory Config. From the src directory:
$ Config/gmh.pl -Av make.install.log
outputs the build status as
--------
Packages
--------
>>>> Fatal Error .... Ignored Error
[ok] Config
[ok] GMAO_Shared
[ok] | GMAO_mpeu
[ok] | | mpi0
[ok] | GMAO_pilgrim
[ok] | GMAO_gfio
[ok] | | r4
[ok] | | r8
[ok] | GMAO_perllib
[ok] | MAPL_cfio
[ok] | | r4
[ok] | | r8
[ok] | MAPL_Base
[ok] | | TeX
[ok] | GEOS_Shared
[ 1] .... .... Chem_Base
[ok] | Chem_Shared
[ok] | GMAO_etc
[ok] | GMAO_hermes
[ 2] .... .... GFDL_fms
[ok] | GEOS_Util
[ok] | | post
-------
Summary
-------
IGNORED mpp_comm_sma.d mpp_transmit_sma.d Chem_AodMod.d (3 files in 2 packages)
All 22 packages compiled successfully.
In case of errors, gmh summarizes exactly where it happens by indicating the package where it occured. Caveat: it does not work in parallel (output is scrambled). So, if the parallel build fails, rerun it sequentially (it will go quickly and die in the same place) and run gmh on the output for a summary.
Advanced features
Check load balance of build
The model includes useful tools like build timers, Config/esma_timer.sh, Config/esma_tgraph.pl. These are useful to time the build and check the load balance of the build, process. There is a way to hook these timers to the build process by setting
ESMA_base.mk:ESMA_TIMER = # command to time build steps (for compilation) ESMA_base.mk:ESMA_TIMER_CI = # command to start timer (for user to backet code segments) ESMA_base.mk:ESMA_TIMER_CO = # command to end timer (for user to backet code segments)
Customize build
A build can be customized by using $HOME/.esma_xxxx.mk
ESMA_BASE = ESMA_base.mk $(wildcard $(HOME)/.esma_base.mk) ESMA_ARCH = ESMA_arch.mk $(wildcard $(HOME)/.esma_arch.mk) ESMA_POST = ESMA_post.mk $(wildcard $(HOME)/.esma_post.mk)
These effectively let you change whatever you want - useful for debugging, etc. For example, you can set your timers in ~/.esma_base.mk.
Run the model
Setting up an experiment
Go into the model application directory and run the setup script:
$ cd $ESMADIR/src/Applications/GEOSgcm_App $ ./gcm_setup
and provide required answers.
MAT: Add a sample run here
Run a one-day test
To run the job, first change to the experiment directory you specified in the above gcm_setup. Then, copy a set of restarts to model directory. Sample restarts are provided on Discover (at /discover/nobackup/mathomp4/Restarts-G10) and Pleiades (at /nobackup/gmao_SIteam/ModelData/Restarts-G10/).
Then, edit CAP.rc:
JOB_SGMT should be 1 day (00000001 000000) NUM_SGMT should be 1
and gcm_run.j by inserting "exit" after
$RUN_CMD $NPES ./GEOSgcm.x set rc = $status echo Status = $rc
A script that can automate these processes is:
#!/bin/bash
sed -i'.orig' -e '/^JOB_SGMT:/ s/000000[0-9][0-9]/00000001/' \
-e '/^NUM_SGMT:/ s/[0-9][0-9]*/1/' \
-e '/^MAPL_ENABLE_TIMERS:/ s/NO/YES/' CAP.rc
sed -i'.orig' -e '/^echo Sta/ a exit' \
-e '/^#PBS -j oe/ a #PBS -m abe\n#PBS -M email@myserver.com' \
-e '/^ if(-e $EXPDIR\/$rst ) \/bin/ s/cp/ln -s/' gcm_run.j
Note: This script also sets up email notification as well as linking restarts to the scratch/ directory rather than copying (which is nice if running at high-res).
Compare to a "correct" run by using cmpscratch
a. If all success, runs are bit-identical b. If not => "What is correct?"