2023-09-05 Render & Cycles Meeting

Attendees

  • Brecht Van Lommel (Blender)
  • Thomas Dinges (Blender)
  • Xavier Hallade (Intel)
  • Nikita Sirgienko (Intel)
  • Patrick Mours (NVIDIA)
  • Brian Savery (AMD)
  • Christophe Hery (Meta)
  • Michael Jones (Apple)

Notes

  • Hydra MaterialX export pull request was submitted, Brecht will review. This exports Blender shader nodes to MaterialX. Native support for MaterialX node editing is not planned for 4.0. Some minor changes planned to the PR to avoid doing premature optimization in export.
  • Brian plans to write a code.blender.org blog post about Hydra, and contact other renderer developers regarding using Hydra.
  • MetalRT curve support pull request was updated by Michael, Brecht will review. Should be straightforward, but actually enabling this requires an upgrade to Xcode 15, where we are getting linker warnings which are proving difficult to fix. We might have to live with those until either Xcode or CMake has a solution to silence them.
  • Michael plans to investigate Metal animation rendering memory leaks, tightening up related code.
  • HIP-RT for Linux is waiting on the release of ROCm 5.7, unclear if this will be in time for 4.0.
  • OpenImageDenoise 2.0 pull request is being reviewed by Sergey. The initial PR brings GPU support for Intel GPUs, with NVIDIA and AMD support coming in a follow up PR. Ideally this would still be in 4.0, but time is short and this requires work from the platform maintainers. Brecht will discuss with Sergey.
  • Shader node previews need some polishing still and are marked experimental, unclear yet if it will be in 4.0 or 4.1.
  • The light linking user interface still needs some polishing for Blender 4.0.
  • Thomas mentioned there are 3 high priority bugs. Brecht will check on them, at least one can have priority lowered.
  • Christophe found a NaN in the new hair BSDF, will send a patch that Brecht will forward to Weizhen.

Practical Info

This is a weekly video chat meeting for planning and discussion of Blender rendering development. Any contributor (developer, UI/UX designer, writer, …) working on rendering in Blender is welcome to join and add proposed items to the agenda.

For users and other interested parties, we ask to read the meeting notes instead so that the meeting can remain focused.

10 Likes

OPTIX 8.0 was released last month. It includes the new ‘Shader Execution Reordering’ feature that improves performance of path tracing by 20-50% according to Nvidia’s white paper. Is an OPTIX upgrade being considered anytime soon?

7 Likes

Cycles already does its own shader based reordering. So the potential performance benefits of this is are difficult to estimate, and numbers from that white paper are not applicable. It may require deep changes, and we may end up with different kernels for OptiX and all the other GPU backends, which is not easy to maintain.

I think there is a lot of potential here, but it’s likely a big project. So upgrading the OptiX version is not really the issue.

On a technical level, what this may entail is having a megakernel with callable programs to keep its size under control, with the state on the stack instead of the heap.

14 Likes

Out of curiosity, is there any new news about Path guiding?

No there is not, it’s something Intel has to implement, and AFAIK there is not defined timeline.
Let’s hope this gets implemented soon, otherwise is like not having it :confused:

1 Like

Just making sure we’re all on the same page. Cycles has Path Guiding (implemented via OpenPGL from Intel) and it currently guides rays for volumetric shaders, and diffuse and rough glossy surfaces. The main limitation at the moment is that it is only supported on the CPU. GPU support is planned, but as @JuanGea mentioned, this will be handled by Intel and there’s no public timeline for it at the moment (at least to my knowledge).

CC @LemonBranny

5 Likes

I assumed that the question was related to GPU PG, and my answer was effectively for that :slight_smile:

1 Like