Scientific App Point of Contacts

[Stephanie Labasan]

We were previously told that the SCC organizers would provide point of contacts for the four scientific applications. When will this information be provided? We are particularly interested in one for Pflotran, since there is very little information publicly available. Also interested in one for Cam, so we can determine which version to use. 

Sincerely,

Stephanie Labasan

University of the Pacific 

[W. Michael Brown]

The committee for the apps includes myself and:

LAMMPS: Trung Dac Nguyen
QMCPack: Jeongnim Kim
CAM: Matt Norman
PFLOTRAN: Richard Mills, Glenn Hammond

possibly others if additional help is needed due to tight schedules.

Please use this mailing list to get help or ask questions for the
scientific apps.

Thanks. - Mike

> --
>
>

[Richard Tran Mills]

Hello All,

Apologies for the lack of information on PFLOTRAN so far. I am one of the
PFLOTRAN developers and the point of contact for the code.

The code should be obtained from the BitBucket.org PFLOTRAN repository at

https://bitbucket.org/pflotran/pflotran-dev

There is a "get source" link you can use at the webpage to download a
tarball or zip archive, but I recommend using the Mercurial
(http://mercurial.selenic.com/) distributed source control tool to obtain
the code, as this will make it easier to update the code. (We have not
yet chosen a revision to "freeze" at for the competition, and it we may
push some changes in response to issues that SCC12 students have.)

Once you have obtained the code, there are some brief installation
instructions available on the wiki on the PFLOTRAN BitBucket site. There
is also a "quick guide" manual for the code that is a work in progress,
which you will find in the PFLOTRAN source repository ('pdflatex' the
quick_guide.tex file in docs/user_manual).

There are lots of PFLOTRAN example problems available in the repository,
but I believe that the most relevant one to experiment with is in the
'shortcourse/exercises/5-spot/' directory. This problem exercises both
the flow and reactive transport capabilities in the code. It also has a
LaTeX file (simple_5spot.tex) that you can build in that directory that
explains what the problem does and shows some visualizations of the
simulation produced with VisIt (https://wci.llnl.gov/codes/visit/).

A few notes about the 5-spot example problem:

* It requires the 'hanford.dat' chemistry database that is in the
'database' directory of the repository. A relative path is specified in
the 'DATABASE' line of the pflotran.in file. If you run the example file
from somewhere else,y you'll need to update the 'DATABASE' line to point
to the correct path.

* The input file is set up to run a problem on a 30 x 30 x 1 grid. You
can easily change the grid resolution by editing the 'NXYZ' line. (Though
I note that I just tried this and get an error for some grid sizes that I
can't immediately explain. I will look into this and get back to people
about this soon.)

Please direct questions that arise to this SCC12 mailing list so that all
participants can benefit from them.

Best regards,
Richard

> --
>
>


--
Richard Tran Mills, Ph.D.
Computational Scientist | Joint Assistant Professor
Computational Earth Sciences Group | Department of EECS
Oak Ridge National Laboratory | University of Tennessee, Knoxville
E-mail: r...@utk.edu Phone: 865-241-3198 http://climate.ornl.gov/~rmills

[Stephanie Labasan]

Hi Richard, 

Thanks so much for all the information regarding Pflotran. I was also wondering if you could provide a sample output for the 5-spot example problem to ensure I'm getting the same. Also, do you have any performance data for Pflotran on any platforms? 

Thanks!

Sincerely,

Stephanie Labasan

University of the Pacific 

[Richard Tran Mills]

Hi Stephanie,

The output from running this example problem is probably a bit on the large side for sending directly to this mailing list, though I could put the output from running this example somewhere.  (Michael, is there a web page where posting this would be appropriate?)  I don't think that the raw output file is necessarily what you want for verifying the correctness of your run, however.  It is hard to make sense of the raw numbers and a plain 'diff' might not work because your results may not match mine bit-for-bit -- the equations are being solved iteratively to within a given tolerance and there might be slight differences in low-order digits between different platforms.  (You could certainly do some sort of "numerical diff" that uses a tolerance for comparison, however.  It might be a useful exercise to use the h5py library to write a Python script to open the pflotran.h5 output files and do this comparison.)

I think it would most be useful for you and the other students to try visualizing the results using the open source, parallel visualization tool VisIt.  You can obtain it from 'https://wci.llnl.gov/codes/visit/' and there is pretty good documentation and some helpful tutorial materials available there.  You will want to load the 'pflotran.h5' file that is produced by your run (use the drop down menu for "Open file as type:" to select 'PFLOTRAN') and then produce pseudocolor plots/movies of 'Chrysocolla2_vf' (volume fraction of the mineral chrysocolla) and pH.  The movies should look like the ones shown in the MPEG files in the 'figs' subdirectory of the 5-spot directory.  These are also shown in the LaTeX write-up: run 'pdflatex simple_5spot.tex' and the movies will be linked into the PDF file.

I note that the 5-spot example file is also configure to dump ASCII Tecplot files (the "pflotran-*.tec" files).  Tecplot is a commonly used commercial plotting tool that your university may have a license for.  Tecplot is more feature-rich but doesn't run in parallel, so it's not suitable for visualizing large simulations.

Best regards,
Richard

--
You received this message because you are subscribed to the Google Groups "SCC12" group.
To post to this group, send email to sc...@googlegroups.com.
To unsubscribe from this group, send email to scc12+un...@googlegroups.com.
Visit this group at http://groups.google.com/group/scc12?hl=en.
 
 

[Stephanie Labasan]

I have a question on the PETSc dependency. I see that there's a new version of PETSc that is 3.3 as opposed to the 3.2-p7. If i build the newer PETSc and run it with Pflotran from "hg clone https://bitbucket.org/pflotran/pflotran-dev", I'm running into the following error: 

$HOME/pflotran/pflotran/src/pflotran/unstructured_grid.F90:4523: undefined reference to `matseqaijgetarray_'

$HOME/pflotran/pflotran/src/pflotran/unstructured_grid.F90:4533: undefined reference to `matseqaijgetarray_'

I was able to get around this error by setting the environment variable PETSC_DIR to point to the older PETSc build v3.2 and including "petsc_api_3_3=1" in my make command, but I'm not sure if that's what we're needing to do. Is there another fix to get around this error and only use v3.3? 

[Richard Tran Mills]

Hi Stephanie,

The tip of the 'pflotran-dev' repository is kept in sync with the development branch of PETSc.  Because the PETSc development team does not hesitate to change the API to make it more intuitive/consistent, there are often changes to the API between releases.  This means that PFLOTRAN is not guaranteed to compile with any PETSc release versions, though we attempt to put the correct #ifdefs in there to support the API of the most recent PETSc release.

Teams should be building PFLOTRAN against the latest petsc-dev.  See

  http://www.mcs.anl.gov/petsc/developers/index.html#obtaining

for instructions on how to obtain it.  Build petsc-dev and update to the latest PFLOTRAN tip ('hg pull -u') and try rebuilding.  If you still encounter issues, please let me know.

Best regards,
Richard

--

You received this message because you are subscribed to the Google Groups "SCC12" group.
To post to this group, send email to sc...@googlegroups.com.
To unsubscribe from this group, send email to scc12+un...@googlegroups.com.
Visit this group at http://groups.google.com/group/scc12?hl=en.
 
 

-- 
Richard Tran Mills, Ph.D.
Computational Earth Scientist      | Joint Assistant Professor
Hydrogeochemical Dynamics Team     | EECS and Earth & Planetary Sciences
Oak Ridge National Laboratory      | University of Tennessee, Knoxville
E-mail: rmi...@ornl.gov  V: 865-241-3198 http://climate.ornl.gov/~rmills

[zhouwe...@gmail.com]

在 2012年9月27日星期四UTC+8上午12时21分00秒，rtm写道：

Hello,Richard

Thanks for your information, I have run some dataset of pflotran, and found that some problem may output a large data in *.tec and .h5 ,sometimes may be 40G+, did I have to save the *.tec and *.h5 and hand in them, or only hand in the *.out?

Simcerely
ZhouWenhao