search

RenderKit / ospray

An Open, Scalable, Portable, Ray Tracing Based Rendering Engine for High-Fidelity Visualization
http://ospray.org
Apache License 2.0
982 stars 178 forks source link

Non overlapped node in AMR #202

Open jakob0 opened 6 years ago

jakob0 commented 6 years ago

Hi! I'm trying to use the AMR structure to visualize our data. I'm impressed by the speed and very happy to finally have some promising tool that fits the needs of our data-structure. Most of the time, it is working well but with some set-ups, I always run into the error: "ERROR: found non overlapped node in AMR structure". When I debug this in the file AMRAccel.cpp, the program finds a possibleSplit not exactly at the bounds, but within the floating point precision. When I add a small tolerance to the comparison of the floating values in the if-statements in lines 96-106, the error is gone for me. Can this be done or should the values always be exactly the same and I am just using it in a wrong way (I wouldn't be surprised, since I'm just beginning to understand your code)?

Thanks Jakob

carsonbrownlee commented 6 years ago

hi Jakob, In your other issue Ingo sent you an email. Were you able to resolve your issue or is this still a problem? We have had a limited set of AMR data to test with and it's unsurprising that issues like this may crop up.

awehrfritz commented 4 years ago

I ran into the same issue:

ERROR: found non overlapped node in AMR structure

Has there been any progress on getting this resolved?

I tested this with the latest ParaView version 5.8 (using the official binary builds for Linux) and the recently added AMReX reader for a dataset (generated by PeleLM). In my setup, I run pvserver on a remote cluster and the ParaView client locally on my laptop. Running pvserver via MPI or in serial mode has no effect on this.

I haven't tested this with simpler dataset, but there are now AMReX test datasets in VTK which should help in debugging this: https://gitlab.kitware.com/vtk/vtk/-/tree/master/Testing/Data/AMReX