/usr/lpp/pessl on all SP nodes.
/usr/lpp/pessl/doc/postscript/pessl.00.ps
PATH:
/afs/ovpit.indiana.edu/@sys/gnu/bin /afs/ovpit.indiana.edu/@sys/X11R6.4/bin /afs/ovpit.indiana.edu/@sys/teTeX/bin /afs/ovpit.indiana.edu/@sys/HDF/bin /afs/ovpit.indiana.edu/@sys/netpbm/binYou should also configure
MANPATH in the same order:
/afs/ovpit.indiana.edu/@sys/gnu/man /afs/ovpit.indiana.edu/@sys/X11R6.4/man /afs/ovpit.indiana.edu/@sys/teTeX/man /afs/ovpit.indiana.edu/@sys/HDF/man /afs/ovpit.indiana.edu/@sys/netpbm/manHaving done all that and having sourced the new environment type:
$ ghostview /usr/lpp/pessl/doc/postscript/pessl.00.ps
PATH:
/usr/afsws/binand then add
/usr/afsws/manto your
MANPATH too. Use the command
$ klog `whoami`to obtain AFS tokens. Use the command
$ tokensto inspect the state of your AFS tokens. They will expire automatically after 25 hours. To destroy your AFS tokens issue the command:
$ unlogNormally you should
unlog explicitly from the AFS before
disconnecting from your UNIX or NT session - that is, unless
you want the tokens to live for, say, background jobs.
If you have problems remembering your AFS password, use the
command:
$ kpasswdto change it. You can change your OVPIT AFS password from any machine (around the world, sic!) that mounts AFS.
/usr/lpp/pessl/example/hpf
gustav@sp20:../SP 16:33:45 !527 $ cd /usr/lpp/pessl/example
gustav@sp20:../example 16:33:51 !528 $ cp -r hpf ~/diffusion
gustav@sp20:../example 16:33:59 !529 $ cd ~/diffusion
gustav@sp20:../diffusion 16:34:05 !530 $ make
xlhpf -c -qreport=hpflist -C -O param.f
** parameters === End of Compilation 1 ===
1501-510 Compilation successful for file param.f.
xlhpf -c -qreport=hpflist -C -O diffusion.f
** diffusion === End of Compilation 1 ===
1501-510 Compilation successful for file diffusion.f.
xlhpf -c -qreport=hpflist -C -O fourier.f
** fourier === End of Compilation 1 ===
1501-510 Compilation successful for file fourier.f.
xlhpf -c -qreport=hpflist -C -O main.f
** main === End of Compilation 1 ===
1501-510 Compilation successful for file main.f.
xlhpf -o diffusion main.o param.o diffusion.o fourier.o \
-lblacs -lpessl -lpesslhpf -lessl
gustav@sp20:../diffusion 16:34:36 !531 $
rsh, which lives
in /usr/afsws/bin is used to make AFS-authenticated
connections to other nodes when running a parallel job,
we don't have things set up that way on our system.
For this reason, it's best to place your binaries and
data file on a file system that is local to the SP
before a production run. It is also much more efficient
to do it this way, because access to data and binaries
on any remote server is always very slow, compared to
a local access.
[...] whose purpose is to create a linear algebra oriented message passing interface that may be implemented efficiently and uniformly across a large range of distributed memory platforms.The main investigator of the project is Jack Dongarra, and the person who does most of the work is R. Clint Whaley. Both are from the University of Tennessee. For more information about BLACS see:
http://www.netlib.org/blacs/
http://www.netlib.org). For
more information about BLAS see:
http://www.netlib.org/blas/
run.script provided
with the example, but it won't work on our system,
because of the way things are set up (e.g., you can
use the User-Space mode from LoadLeveler jobs only).
.rhosts file
in your home directory. That file should look
as follows:
gustav@sp20:../SP 17:04:09 !576 $ cat .rhosts sp17 sp18 sp19 sp20 sp21 sp22 sp23 sp24 sp40 sp41 sp42 sp43 gustav@sp20:../SP 17:04:23 !577 $These are all the nodes that we are allowed to work with in this course. You can use them for interactive runs, or you can submit your jobs to either the
pa class or to the
test class.
.rhosts
or link to a file called
host.list in your run directory, i.e.,
the directory where you've got your binaries and
your data files:
gustav@sp20:../SP 17:04:23 !578 $ cd diffusion gustav@sp20:../SP 17:04:23 !579 $ ln -s ~/.rhosts host.list gustav@sp20:../SP 17:04:23 !580 $
gustav@sp20:../SP 17:04:23 !580 $ poe diffusion -euidevice css0 \
-euilib ip -procs 8 -ilevel 3 \
> diffusion.log 2> diffusion.err
gustav@sp20:../SP 17:04:23 !581 $
diffusion.log and various
diagnostic messages and/or errors on
diffusion.err.
diffusion.err looks
as follows:
INFO: DEBUG_LEVEL changed from 0 to 1 D1<L1>: Open of file ./host.list successful D1<L1>: mp_euilib = ip D1<L1>: task 0 sp17.ucs.indiana.edu 129.79.7.68 10 D1<L1>: task 1 sp18.ucs.indiana.edu 129.79.7.69 10 D1<L1>: task 2 sp19.ucs.indiana.edu 129.79.7.70 10 D1<L1>: task 3 sp20.ucs.indiana.edu 129.79.7.71 10 D1<L1>: task 4 sp21.ucs.indiana.edu 129.79.7.72 10 D1<L1>: task 5 sp22.ucs.indiana.edu 129.79.7.73 10 D1<L1>: task 6 sp23.ucs.indiana.edu 129.79.7.74 10 D1<L1>: task 7 sp24.ucs.indiana.edu 129.79.7.75 10This means that this job has run successfully on 8 SP nodes, sp17 through sp24.
-ilevel 3 or higher.
To request -ilevel other than the default,
which is 1, or 0 is costly and should not be
normally done for production runs.
diffusion.log yields the normal
output of the program, i.e., tabulated diffusion
results for various points and times.
Our first task will be to replace that with an HDF
animation.
This is what the beginning of that file looks like:
point # X Y
1 .5236 .5236
2 1.0472 .5236
3 1.5708 .5236
4 2.0944 .5236
5 2.6180 .5236
...
21 .5236 2.6180
22 1.0472 2.6180
23 1.5708 2.6180
24 2.0944 2.6180
25 2.6180 2.6180
Points
TIME # 1 # 2 # 3 # 4 # 5 # 6
.10000 1.87754 2.99610 3.37190 2.99610 1.87754 3.02958
.20000 1.61896 2.62389 2.93406 2.55544 1.51137 2.63140
.30000 1.40707 2.29776 2.55618 2.18506 1.25438 2.30026
.40000 1.22871 2.01193 2.22646 1.87602 1.05857 2.01355
.50000 1.07478 1.75988 1.93729 1.61515 .90145 1.76125
...
1.50000 .26483 .42718 .45803 .36980 .20098 .42757
1.60000 .22898 .36914 .39549 .31908 .17332 .36947
1.70000 .19790 .31888 .34143 .27530 .14948 .31917
1.80000 .17098 .27539 .29472 .23753 .12893 .27564
1.90000 .14767 .23778 .25437 .20493 .11121 .23800
2.00000 .12752 .20527 .21952 .17680 .09592 .20546