[Paraview] loading vtu files crashes paraview in parallel

[Mohammad Mirzadeh]

Well then let me explain I'm hoping to achieve. Maybe there is a better to
achieve what I'm trying to do.

I am trying to visualize the parallel partitioning in my application (not
paraview) by moving separate partitions to different locations. Initially I
was doing this by creating threshold filters on the pvtu file that was
pointing to individual vtu files. For some reason PV runs out of memory
when doing that even though I'm doing this on a large number of processors
(100M unstructured grid partitioned among 256 processors and visualized
using paraview on 128 processors).

Anyway since that did not work, I though I would avoid the thresholds
altogether and load individual vtu files manually to have control over
individual pieces. It seems that if I use pvd instead of pvtu I can also
have access over individual parts using ExtractBlock filter but I'm not
sure if that would be any different than using the threshold filter ...
which did not work.

Is there any better way of achieving what I'm trying to do here?


On Wed, Apr 16, 2014 at 11:42 AM, Utkarsh Ayachit <
utkarsh.ayachit at kitware.com> wrote:

> It's unclear if the readers are being deleted and their memory
> released based on what you say and hence not sure what could be
> causing the problem. Another thing to try maybe the following:
>
> + Create a .vtm file that refers all of the vtu files. Attached file
> shows the format. just add <DataSet /> elements for each of the files
> with correct index. Note the file paths are relative to the vtm file.
> + Now open this and contour it. In parallel, ParaView will read each
> vtu file separately on each of the processes.
>
> Utkarsh
>
> On Wed, Apr 16, 2014 at 1:00 PM, Mohammad Mirzadeh <mirzadeh at gmail.com>
> wrote:
> > When does the distribution happen?
> >
> > I have a loop in which I load individual vtu files, then create a contour
> > and then go to next vtu file. After all the files are load3d and the
> > contours created I issue a render and that when PV crashes.
> >
> > Also if I run pvbatch with --symmetric option, can I directly load vtu
> files
> > on each process and work with the contours in parallel?
> >
> > On Apr 16, 2014 6:51 AM, "Utkarsh Ayachit" <utkarsh.ayachit at kitware.com>
> > wrote:
> >>
> >> For *.vtu files, ParaView reads the data on the 0th node and then
> >> partitions and redistributes it to all ranks.
> >>
> >> On Tue, Apr 15, 2014 at 7:00 PM, Mohammad Mirzadeh <mirzadeh at gmail.com>
> >> wrote:
> >> > just asking this question in a different way. What happens if I load a
> >> > vtu
> >> > file in parallel? Can paraview handle that automatically (i.e.
> >> > partitioning
> >> > it and sending it to different processors) or is every processor going
> >> > to
> >> > load the same exact file?
> >> >
> >> >
> >> > On Thu, Apr 10, 2014 at 6:10 PM, Mohammad Mirzadeh <
> mirzadeh at gmail.com>
> >> > wrote:
> >> >>
> >> >> Hi,
> >> >>
> >> >> I have a modest size problem, 100M unstructured, and I can easily
> load
> >> >> into paraview with 64 processors / 128G ram when I load the pvtu
> file.
> >> >>
> >> >> I can also work on the file, meaning creating contours, etc. However,
> >> >> when
> >> >> I load all vtu files separately and create contours on each file
> >> >> separately
> >> >> it seems that PV runs out of memory and my process gets terminated.
> >> >>
> >> >> Is this normal?
> >> >>
> >> >
> >> >
> >> > _______________________________________________
> >> > 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
> >> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.paraview.org/pipermail/paraview/attachments/20140416/0951d33b/attachment-0001.html>