Please have a programmer look at this. I dived already deep in the code and fished out the problematic function for you and explained why this is happening. I am not uploading some 20GB data somewhere just to prove what a programmer would understand instantly. Thanks
Basically any code that uses this function will encounter this bug as long as connectivity is an int32 array and length(connectivity) > INT32_MAX. Data triggers this bug, not code.
Crash with OOM due to the fact that the offset array generated from this helper uses the same 32bit structure of connectivity. We have 32bit connectivity array with (point) indices well below INT32_MAX, but more than INT32_MAX cells! Hence the array size is huge and offsets exceed the INT32 limit and fail miserably. Converting the connectivity to 64bit solves the problem but wastes a lot of memory.
Why do you enforce same type for connectivity and offset? Offset could go automatically to 64bit, helper function knows if 32bit are exceeded from size of connectivity.
Best, Ch
From the download section, the Windows and Linux 5.11 versions miss View -> Animation View -> + Camera Orbit. Is this intended or has it moved?
Thanks!
My bug hunt lead me here too. It looks like vtkConduitArrayUtilities do not support strided arrays as conduit does. At least an error should be issued.
Our particle memory layout is X,Y,Z,value,value,value,X,Y,Z,v... which is fine to represent with conduit's set_external but I guess there is no (easy) equivalent for vtkArrays.
I actually have a similar problem, set_external() with linear double arrays work, but strided I get sometimes garbage although using node.print() shows everything in the green, so it must be on the paraview side interpretation of strided conduit arrays. In the pipeline script I get the wrong bounds already... Maybe pv is interpreting the stride settings as elements instead of bytes?
PV 5.9.1 Linux (from downloads), File->Load Path Tracer Materials...
Change something in the file, repeat, nothing happens. Only reloaded if file-name is different (so it seems).