[Paraview-developers] [Non-DoD Source] Re: Change in mpi operation of paraview and pvpython in ParaView 5.5.0-RC1-1

[Andy Bauer]

FYI: I just made an MR for this --
https://gitlab.kitware.com/paraview/paraview-superbuild/merge_requests/440.

I made it an advanced option, also called PARAVIEW_INITIALIZE_MPI_ON_CLIENT,
and only available if mpi is enabled.

On Fri, Mar 9, 2018 at 3:26 PM, Andy Bauer <andy.bauer at kitware.com> wrote:

> Hi Joe,
>
> I don't see PARAVIEW_INITIALIZE_MPI_ON_CLIENT being set in the PV
> superbuild. It does make sense to me though to add that option to the
> superbuild though. As far as the default value for
> PARAVIEW_INITIALIZE_MPI_ON_CLIENT in ParaView itself, I guess it's just a
> matter of opinion on whether it should be enabled or disabled by default.
>
> Best,
> Andy
>
> On Fri, Mar 9, 2018 at 3:13 PM, Hennessey, Joseph G CTR USARMY RDECOM ARL
> (US) <joseph.g.hennessey2.ctr at mail.mil> wrote:
>
>> Andy,
>>
>> Running paraview with --no-mpi remove the requirement for starting MPI
>> with
>> it.
>>
>> It seems that the superbuild is defining PARAVIEW_INITIALIZE_MPI_ON_CLI
>> ENT
>> for some reason.
>>
>> Looking in the RC2 of ParaView 5.5.0's main CMakeLists.txt file
>>
>> configure_file(
>>   ${ParaView_SOURCE_DIR}/vtkPVConfig.h.in
>>   ${ParaView_BINARY_DIR}/vtkPVConfig.h
>>   @ONLY)
>>
>> Is now setting PARAVIEW_INITIALIZE_MPI_ON_CLIENT
>> whereas with ParaView 5.4.1 it was being initialized to off.
>>
>> I used cmake 3.10.2 to configure the superbuild for ParaView 5.5.0-RC2,
>> while I used cmake 3.9.1 to configure the superbuild for ParaView 5.4.1.
>>
>> Thanks,
>>
>> Joe
>>
>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>> Joseph G. Hennessey Ph.D., SAIC
>> Team SAIC
>> Army Research Lab
>> DOD Supercomputing Resource Center
>> Email:  joseph.g.hennessey2.ctr at mail.mil
>>
>>
>> -----Original Message-----
>> From: Andy Bauer [mailto:andy.bauer at kitware.com]
>> Sent: Wednesday, March 07, 2018 9:22 AM
>> To: Hennessey, Joseph G CTR USARMY RDECOM ARL (US)
>> <joseph.g.hennessey2.ctr at mail.mil>
>> Cc: ParaView Developers <paraview-developers at paraview.org>
>> Subject: [Non-DoD Source] Re: [Paraview-developers] Change in mpi
>> operation of
>> paraview and pvpython in ParaView 5.5.0-RC1-1
>>
>>
>> Hi Joe,
>>
>>
>> Is it possible that you're using a different MPI for your builds?  I'm
>> wondering if the previous MPI version didn't require mpirun in order to
>> run a
>> single process while the current one does.
>>
>> What you can do though is to use the "--no-mpi" option which will skip
>> initializing and finalizing MPI but also make the MPI required filters and
>> readers not available. It's possible that the CMake
>> PARAVIEW_INITIALIZE_MPI_ON_CLIENT option got switched from OFF to ON. It
>> it is
>> OFF by default but PV is still built with MPI you can do a --mpi to have
>> the
>> GUI or pvpython initialize and finalize MPI.
>>
>>
>> Most filters (actually, I'm not aware of any) don't explicitly rely on
>> MPI for
>> single process runs but there are a couple of readers that use MPI IO.
>>
>>
>> Best,
>>
>> Andy
>>
>>
>> On Wed, Mar 7, 2018 at 9:02 AM, Hennessey, Joseph G CTR USARMY RDECOM ARL
>> (US)
>> <joseph.g.hennessey2.ctr at mail.mil <
>> Caution-mailto:joseph.g.hennessey2.ctr at mail.mil > > wrote:
>>
>>
>>         Hello,
>>
>>         I have noticed that when building ParaView 5.5.0-RC1-1 with an
>> external
>>         MPI version that pvpython and paraview itself are now being built
>> such
>>         that they can not be executed without mpirun.
>>
>>         For example this would work with ParaView 5.4.1
>>
>>         paraview
>>         pvpython
>>         mpirun -n 1 pvbatch
>>
>>         While for ParaView 5.5.0-RC1-1 I have to do the following
>>
>>         mpirun -n 1 paraview
>>         mpirun -n 1 pvpython
>>         mpirun -n 1 pvbatch
>>
>>         as running
>>
>>         paraview
>>         pvpython
>>
>>         without mpirun will no longer work and complain about
>>         attempting to run an mpi required program without mpi,
>>         as has been the case with pvbatch in the past.
>>
>>         Is this an intentional change, or is this an unintended
>> consequence
>>         of some other change?
>>
>>         Thanks,
>>
>>         Joe
>>
>>         ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>         Joseph G. Hennessey Ph.D., SAIC
>>         Team SAIC
>>         Army Research Lab
>>         DOD Supercomputing Resource Center
>>         Email:  joseph.g.hennessey2.ctr at mail.mil <
>> Caution-mailto:joseph.g.hennessey2.ctr at mail.mil >
>>
>>
>>         _______________________________________________
>>         Powered by Caution-www.kitware.com < Caution-
>> http://www.kitware.com >
>>
>>         Visit other Kitware open-source projects at
>> Caution-http://www.kitware.com/opensource/opensource.html <
>> Caution-http://www.kitware.com/opensource/opensource.html >
>>
>>         Search the list archives at:
>> Caution-http://markmail.org/search/?q=Paraview-developers <
>> Caution-http://markmail.org/search/?q=Paraview-developers >
>>
>>         Follow this link to subscribe/unsubscribe:
>>         Caution-https://public.kitware.com/mailman/listinfo/paraview
>> -developers <
>> Caution-https://public.kitware.com/mailman/listinfo/paraview-developers >
>>
>>
>>
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://public.kitware.com/pipermail/paraview-developers/attachments/20180309/976e3264/attachment.html>