[Paraview] Statically linking catalyst to fortran simulation code on Titan

Fri Sep 13 14:46:03 EDT 2013

Thanks for all the useful info, David, and Andy.

After spending a lot of time looking into Cmake trace aLINK_SEARCH_END_STATICnd FindMPI.cmake, I finally resolved this problem by setting on LINK_SEARCH_START_STATIC and LINK_SEARCH_END_STATIC flags. Now the resulting executable is static and I can run in-situ viz with our simulation on Titan. However, I don't see the slice images generated by Catalyst as set up in the python pipeline although the output message indicated that coprocessing was called and returned with no error. I am going to look into the code next to investigate further, but a quick relevant question in my mind is: is this one static executable sufficient to run on the compute node or I have to copy over python modules and set python path appropriately for the pipeline to be executed by Catalyst?

Thanks and regards,

Hong

________________________________
From: Andy Bauer [andy.bauer at kitware.com]
Sent: Thursday, September 12, 2013 6:07 PM
To: David E DeMarle
Cc: Hong Yi; paraview at paraview.org
Subject: Re: [Paraview] Statically linking catalyst to fortran simulation code on Titan

If I remember correctly, FindMPI.cmake is trying to set the language specific libraries (e.g. MPI_CXX_LIBRARIES) and if those aren't set then it will try to find them and finally overwrite what is in MPI_LIBRARY. So I would suggest changing what Dave D. suggested with MPI_{C,CXX,Fortran}_LIBRARIES set in addition to setting MPI_LIBRARY. If you look at FindMPI.cmake, you'll see that MPI_LIBRARY and MPI_EXTRA_LIBRARY are deprecated.

Andy

On Thu, Sep 12, 2013 at 5:33 PM, David E DeMarle <dave.demarle at kitware.com<mailto:dave.demarle at kitware.com>> wrote:
They are overridden in the CMakeCache then? Ie you preset them and something later changes them.

In situations like that cmake --trace is helpful. It tells cmake to spit out everything it does to the terminal. Pipe that into a file, and then grep the file for the cmake variable in question. You will be able to see what exactly set it incorrectly.

It will also be helpful to take a look at the FindMPI module itself, as that is the most likely culprit. By reading that it might become apparent.

good luck!

David E DeMarle
Kitware, Inc.
R&D Engineer
21 Corporate Drive
Clifton Park, NY 12065-8662
Phone: 518-881-4909<tel:518-881-4909>

On Thu, Sep 12, 2013 at 5:26 PM, Hong Yi <hongyi at renci.org<mailto:hongyi at renci.org>> wrote:
Thanks for the info, David. I did try setting all these flags you indicated when building simulation code and Catalyst Fortran example code with CMake, but it does not seem to help. Let me know if you have any other suggestions.

Regards,
Hong

From: David E DeMarle [mailto:dave.demarle at kitware.com<mailto:dave.demarle at kitware.com>]
Sent: Thursday, September 12, 2013 5:03 PM
To: Hong Yi
Cc: paraview at paraview.org<mailto:paraview at paraview.org>
Subject: Re: [Paraview] Statically linking catalyst to fortran simulation code on Titan

I preset these config flags to make sure I got the static MPI libs.

list(APPEND PARAVIEW_OPTIONS "-DBUILD_SHARED_LIBS:BOOL=OFF")
list(APPEND PARAVIEW_OPTIONS "-DPARAVIEW_USE_MPI:BOOL=ON")
list(APPEND PARAVIEW_OPTIONS "-DMPI_INCLUDE_DIRS:STRING=/opt/cray/mpt/5.5.5/gni/mpich2-gnu/47/include")
list(APPEND PARAVIEW_OPTIONS "-DMPI_LIBRARY:FILEPATH=/opt/cray/mpt/5.5.5/gni/mpich2-gnu/47/lib/libmpich.a")
list(APPEND PARAVIEW_OPTIONS "-DMPI_EXTRA_LIBRARY:FILEPATH=/opt/cray/mpt/5.5.5/gni/mpich2-gnu/47/lib/libmpichcxx.a")
list(APPEND PARAVIEW_OPTIONS "-DVTK_USE_SYSTEM_HDF5:BOOL=1")

It is likely the case that presetting these in your simulation code will make it choose the right ones too:
list(APPEND PARAVIEW_OPTIONS "-DMPI_INCLUDE_DIRS:STRING=/opt/cray/mpt/5.5.5/gni/mpich2-gnu/47/include")
list(APPEND PARAVIEW_OPTIONS "-DMPI_LIBRARY:FILEPATH=/opt/cray/mpt/5.5.5/gni/mpich2-gnu/47/lib/libmpich.a")
list(APPEND PARAVIEW_OPTIONS "-DMPI_EXTRA_LIBRARY:FILEPATH=/opt/cray/mpt/5.5.5/gni/mpich2-gnu/47/lib/libmpichcxx.a")

David E DeMarle
Kitware, Inc.
R&D Engineer
21 Corporate Drive
Clifton Park, NY 12065-8662
Phone: 518-881-4909<tel:518-881-4909>

On Thu, Sep 12, 2013 at 4:57 PM, Hong Yi <hongyi at renci.org<mailto:hongyi at renci.org>> wrote:
I am able to successfully build our Fortran simulation code as well as Catalyst Fortran example code linked to catalyst (built with superbuild), but the resulting executable is not static, but have some dependencies on shared libraries such as libmpichf90_gnu_47.so.1, limxpmem.so.1, etc., which seems to be resulting from “find_package(MPI)” in CMakeLists.txt that somehow find shared mpi libraries to link to. This dependencies prevents the executables from running on Titan compute nodes. Since paraview superbuild cross target can generate pvserver as well as other executables statically without any dependencies on other shared libraries, I have been trying to study paraview superbuild trying to build Catalyst in-situ Fortran example code also statically without any dependencies, but have not been able to succeed. I guess one way to do it is to manually specify all those MPI static libraries in CMake, however, I have not been able to succeed due to so many dependencies of those MPI libraries. Could anyone advise on how to achieve static linking to MPI libraries automatically by CMake via find_package(MPI)?

Thanks,
Hong

_______________________________________________
Powered by www.kitware.com<http://www.kitware.com>

Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html

Please keep messages on-topic and check the ParaView Wiki at: http://paraview.org/Wiki/ParaView

Follow this link to subscribe/unsubscribe:
http://www.paraview.org/mailman/listinfo/paraview

_______________________________________________
Powered by www.kitware.com<http://www.kitware.com>

Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html

Please keep messages on-topic and check the ParaView Wiki at: http://paraview.org/Wiki/ParaView

Follow this link to subscribe/unsubscribe:
http://www.paraview.org/mailman/listinfo/paraview

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.paraview.org/pipermail/paraview/attachments/20130913/e29bc3c0/attachment.htm>