[Paraview] Adding not rendered filters slows down animation

Utkarsh Ayachit utkarsh.ayachit at kitware.com
Thu Jul 12 10:31:10 EDT 2018


BTW, I've reported the reexecution of the Threshold1 geometry
generation stage as a bug here:
https://gitlab.kitware.com/paraview/paraview/issues/18293

Utkarsh
On Thu, Jul 12, 2018 at 10:25 AM Utkarsh Ayachit
<utkarsh.ayachit at kitware.com> wrote:
>
> Ole,
>
> That indeed doesn't sound like expected behavior. A few questions
> before we can diagnose what's going on: when you say "adding a second
> threshold to the first slows things down heavily, even though
> threshold 2 is not rendered", can you explain what do you mean by
> that? Here what I did: selected the Threshold1 in the PIpeline Browser
> and created nother Threshold (leaving values as default). I haven't
> hit Apply yet, since as soon as I do, ParaVIew will show Threshold2's
> output. At this point, I don't see any change in rendering/update
> times. If I hit "Apply", Threshold2 executes and hence the Update
> takes time, as expected.  At this point, if I hide "Threshold2" and
> turn on "Threshold1", the first render is a little slower since the
> representation is regenerating the geometry for Threshold1 -- that's
> indeed a bug and we'll track that down. Subsequent renders, however,
> seem unaffected. What am I doing wrong?
>
> Utkarsh
> On Thu, Jul 12, 2018 at 5:11 AM Ole Numssen <numssen at cbs.mpg.de> wrote:
> >
> > Hello there,
> >
> > I'm experiencing some strange performance/timing differences when creating animations with paraview 5.5.2 on Ubuntu 16.04 for different filter combinations.
> >
> > I think I've broken it down to a minimal example. It seems to be of importance, if the rendered Filter has a child-filter or not, independent of this child filter being rendered.
> > I've taken some timings for a 2-frame animation to make things clear. I've attached the complete Timer Log output for the 4 different setups shown below.
> >
> > Filters shown as in the pipeline browser. I've loaded an .xdmf/.hdf5 file, Multi-block and Multi-pice Dataset, ~8.5M Cells, ~2MPoints on a i7-4770K CPU @ 3.50GH 4 core machine, 24G Ram, Titan XP gpu.
> >
> > 1) quite fast at first
> > file.xdmf (not rendered)
> > | Threshold 1 (rendered)
> >
> > -> Frame 1: 0.00056 seconds
> > -> Frame 2: 0.000385 seconds
> >
> > 2) adding a second threshold to the first slows things down heavily, even though threshold 2 is not rendered.
> > file.xdmf (not rendered)
> > | Threshold 1 (rendered)
> > -| Threshold 2 (not rendered)
> >
> > -> Frame 1: 10.1624 seconds
> > -> Frame 2: 10.1568 seconds
> >
> >
> > 3) deleting threshold 2 speeds things up a little (only frame 2)
> > file.xdmf (not rendered)
> > | Threshold 1 (rendered)
> > -| Threshold 2 (not rendered)
> >
> > -> Frame 1: 0.000553 seconds
> > -> Frame 2: 9.87154 seconds
> >
> >
> > 4) deleting threshold 1 and creating it again leads to fast animation again:
> > file.xdmf (not rendered)
> > | Threshold 1 (rendered)
> >
> > -> Frame 1: 0.000475 seconds
> > -> Frame 2: 0.000489 seconds
> >
> >
> > Is this the intended behavior? At least for me it is not the expected behavior, as the rendered image/data is exactly the same for all of the 4 states shown above.
> >
> > Cheers,
> > Ole_______________________________________________
> > Powered by www.kitware.com
> >
> > ParaView discussion is moving! Please visit https://discourse.paraview.org/ for future posts.
> >
> > 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
> >
> > Search the list archives at: http://markmail.org/search/?q=ParaView
> >
> > Follow this link to subscribe/unsubscribe:
> > https://public.kitware.com/mailman/listinfo/paraview


More information about the ParaView mailing list