[Fwd: [Paraview] Run Paraview sort-last on several graphics
devices in the same machine]
Qinghuai Gao
Qinghuai.Gao at sun.com
Thu Nov 17 20:38:51 EST 2005
Ken:
I come back to this problem again. I am trying to set the DISPLAY in
the code at MPI_Init. However, I am not sure this really does the work
as expected because Paraview has a very complicated architecture.
Precisely, I add a paragraph after the MPI_Init in
VTK/Parallel/vtkMPIController.cxx. Please confirm me this is the right
place to add, or point me where is the right place if this is not.
I can verify that the data is processed by two process by adding a
"Process Id Scalar" filter.
However, how can I verify the render task is assigned to two graphic
cards. Any thoughts?
Thanks,
Qinghuai
> I have never used a machine with many graphics devices, but I can tell
> you that to get parallel rendering to work you will need to run ParaView
> in some kind of MPI mode. The actual mode you use (local,
> client/server, client/render-server/data-server) depends on your setup,
> but they are all capable of parallel rendering.
>
> Getting each process in your MPI job to use a different graphics device
> may be more tricky. Offhand, I can't think of an easy way to do it.
> Perhaps you could have mpirun (or whatever launches your MPI jobs) call
> a script that somehow determines the process' rank, sets the DISPLAY
> accordingly, and then calls the real ParaView implementation.
>
> -Ken
>
>
>>-----Original Message-----
>>From: paraview-bounces+kmorel=sandia.gov at paraview.org
>>[mailto:paraview-bounces+kmorel=sandia.gov at paraview.org] On
>>Behalf Of Qinghuai Gao
>>Sent: Tuesday, September 20, 2005 11:59 AM
>>To: paraview at paraview.org
>>Subject: [Fwd: [Paraview] Run Paraview sort-last on several
>>graphics devices in the same machine]
>>
>>I asked this question last week but got no reply yet. Wonder
>>anyone try this kind configuration to run Paraview with many
>>graphics devices in a single machine. If so, please shine
>>some light. Thanks in advance,
>>
>>Qinghuai
>>
>
>
More information about the ParaView
mailing list