Ok - that clears it up. I will give that a try.
I understand what you're saying wrt the small size of my parallel PV
setup. I am on a single workstation trying to use the fact I have many
cores to speed things up a bit. It was suggested in a previous
question to the (from [email protected] on Aug 16, [Paraview]
running in parallel on OS X) list that I'd want to stick with just 2
or 4 CPUs to minimize contention for disk and GPU rendering time.
Thanks again.
Dan
On 31-Aug-10, at 9:55 AM, Moreland, Kenneth wrote:
This will not prevent pvserver from getting involved in the
reading. pvserver always reads in the data. That’s just the way
ParaView works.
If the reader is smart (and I believe the XDMF reader is), then it
will read the partition on process 0 of pvserver, and process 1 will
simply generate an empty data set.
BTW, a two process pvserver is not very big. Going parallel
involves overhead with data management and communication. Thus for
some operations you are not going to see a 2X speed improvement.
-Ken
On 8/31/10 8:10 AM, "Dan Lussier" <[email protected]> wrote:
Ok - thanks. Good to know I'm on the right track.
So to confirm:
- I have a GUI running and 2 pvserver processes running on a single
workstation.
- I am able to connect to the servers
- I try to open the file as I would normally do on the GUI.
- I then run the D3 filter.
How is this approach going to prevent the pvserver processes from
getting involved in the reading?
For reference my data is XDMF Polyvertex (i.e. unstructured) grid.
Dan
On Mon, Aug 30, 2010 at 4:38 PM, Moreland, Kenneth
<[email protected]> wrote:
> I think you’ve answered your own question. Read the data into
ParaView.
> Run the D3 filter on it. The output of the D3 filter will be
distributed
> amongst the processes.
>
> -Ken
>
>
> On 8/30/10 3:23 PM, "Dan Lussier" <[email protected]> wrote:
>
> I am trying to start using Paraview in parallel for the first time
and
> have run into some trouble reading my data in when running in
> parallel.
>
> I am using XDMF and have unstructured data, which I have gathered
from
> the mailing list (thanks!) cannot be read in parallel without
> restructuring the XML to use hyperslabs.
>
> In scanning a set of slides from an online tutorial I saw the
> following method suggested if you have a reader that isn't parallel:
>
> • Only one node reads
> • Whole data set must fit in one process’ memory
> • If you must, the D3 filter in Paraview will partition and
> distribute
>
> How would I go about getting the data read in by just one process
and
> then distributing it to other processors after read is complete?
>
> Sorry if this question is a bit off the wall but I'm open to
suggestions.
> _______________________________________________
> Powered by 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
>
>
>
>
> **** Kenneth Moreland
> *** Sandia National Laboratories
> ***********
> *** *** *** email: [email protected]
> ** *** ** phone: (505) 844-8919
> *** web: http://www.cs.unm.edu/~kmorel
>
>
**** Kenneth Moreland
*** Sandia National Laboratories
***********
*** *** *** email: [email protected]
** *** ** phone: (505) 844-8919
*** web: http://www.cs.unm.edu/~kmorel
_______________________________________________
Powered by 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
