I was just doing some testing with 6.2.0 with the official Linux binaries and noticed that only the 2 OSPRay backends are now available, and the OptiX option is gone? Yet the latest manual still describes it.
And while looking at the available ANARI options it seems to be based on visrtx (based on warnings about unsupported features), which thankfully seems to be included in the binaries. But is missing from the docs.
The OSPRay pathtracer seems to be broken, as increasing samples per pixel turns the rendered object blue (and I vaguely remember seeing this before, but can’t find any bug report):
So at this point I’m kind of wondering what to expect when it comes to raytraced rendering in 6.2.0, are these known issues? Am I doing something wrong?
Thank you for the thorough write-up, Paul. You are not doing anything wrong.
The missing “OptiX pathtracer” option is a side-effect of introducing ANARI with VisRTX. To ship the new VisRTX with ParaView, the legacy OptiX-based VisRTX library had to be dropped to avoid having two conflicting VisRTX libraries shipped with ParaView. There is some ongoing investigation to see if we can keep providing the legacy OptiX backend until the ANARI render path stabilizes within ParaView.
When ANARI is enabled, ParaView selects the backend as follows: if ANARI_LIBRARY is set, ParaView honors that selection; otherwise, it defaults to visrtx on Linux/Windows and helide on macOS. There is also a work-in-progress branch for ParaView here: https://gitlab.kitware.com/paraview/paraview/-/merge_requests/7980. It displays the currently selected ANARI backend and provides a drop-down selector for switching between available backends.
With regards to the background color in ANARI VisRTX “quality” mode, I think it is either a bug in ParaView or the version of VisRTX that was shipped with ParaView 6.2.0. I verified locally that the background is properly rendered in quality mode with the latest anari-sdk and visrtx. See the image below. I’ll dig into this further and report back.
I am working on the known-working OSPRay 2.7.1 for ParaView 6.2.1. If anyone has time to debug the issues with OSPRay 2.12, that would be fine, too, but I suspect that won’t be done before the ParaView 6.2.1 expected in early November.
