Building and running the wave example on orcinus
Commands that you will need to type are in bold. It is also assumed
that DV is running on your local workstation, and that in
your orcinus shell you have set DVHOST to the name
of said workstation, e.g.
% setenv DVHOST bh0.phas.ubc.ca
Change to a convenient working directory (perhaps your home) and
copy the example from Matt's templates directory:
% cp -a ~matt/templates/pamr-wave .
Change to pamr-wave, and ensure that the proper compilation
environment is set up (source command)
% cd pamr-wave
% source ~matt/scripts/soINTEL-mpi-orc
CC = /global/software/openmpi-1.4.3/intel//bin/mpicc
LDFLAGS = -O3 ...
Make the application
Wait for the batch job to finish: you can use qstat -u $LOGNAME
command to monitor the status of your job, but you can tell when the
job has completed by the appearance of the standard output and
standard error files associated with the run. These will have
names of the form pamr-wave.o[Job
id] and pamr-wave.e[Job
id] where [Job id]
is the numeric part of the Job id which is returned by the qsub command.
Submit the batch job
% qsub pamr-wave.pbs
Once the run has completed, the directory should contain 8 .sdf files, one for each
of the 8 processors that were used in the sample computation.
% cd run_2d
% ls *sdf
wave_2d_L0_phi_tl2_0.sdf wave_2d_L0_phi_tl2_3.sdf wave_2d_L0_phi_tl2_6.sdf
wave_2d_L0_phi_tl2_1.sdf wave_2d_L0_phi_tl2_4.sdf wave_2d_L0_phi_tl2_7.sdf
Send all of these to DV using sdftodv:
% sdftodv *sdf
Then select Merge -> All Registers in DV, enable
AMR in the Options panel, and visualize the results.
You can use Makefile and pamr-wave.pbs as templates for your own
Maintained by firstname.lastname@example.org.
Supported by CIAR, CFI and NSERC