2023-01-10 Render & Cycles Meeting

Attendees

  • Brecht Van Lommel (Blender)
  • Weizhen Huang (Blender)
  • Thomas Dinges (Blender)
  • Lukas Stockner (Blender)
  • Jeffrey Liu
  • Nikita Sirgienko (Intel)
  • Sebastian Herholz (Intel)
  • Michael Jones (Apple)
  • Patrick Mours (NVIDIA)
  • Christophe Hery (Meta)
  • Brian Savery (AMD)

Blender 3.5

  • Light tree
    • Jeffrey works on fixing bug with sun light sampling. Some discussion on best solution and practical implementation issues.
    • Brecht would still like to do multi-threading and instancing optimizations.
  • AMD HIP:
    • Linux RDNA2 crash bug was confirmed by AMD team and is being investigated.
    • HIP-RT support is still being worked on. Code should be submitted for review even if driver is not out yet.
  • Apple Metal:
    • Tuning will be submitted still for the latest kernel changes.
    • Support for NanoVDB is also expected to land still.
  • OptiX OSL
    • Patrick still plans improvements to better match noise functions with CPU.
    • Brecht also suggested to look at supporting texture file path loading through SVM, even if that would not support all the same filtering options as OIIO.

Other

  • Sebastian found discrepancies between lights that use ray visibility to make them not affect diffuse or glossy surfaces. A patch to address this was submitted, Brecht will review. Some discussion on best approach for this and performance impact.
  • Weizhen has been working on integrating a new Hair BSDF. The biggest challenge right now is for elliptical shaped hair fibers, which raises questions about how best to raytrace curves with such shapes (if at all), how to compute the normal direction for the ellipse, and how to deal with Catmull-Rom not being C2 continuous.
  • Patrick made progress on resumable rendering and denoising, with ability to store and load EXR caches for an animation sequence. Next step is loading cache for previous frame, which requires some work as this denoising code path does not support it yet.
  • Lukas continues work on completing the Principled BSDF v2. The plan is to start merging individual parts, like new multiscatter GGX, new sheen closures in the Velvet BSDF, and then at the end put it all together in the Principled BSDF.
    • Christophe Hery mentioned the Subsurface IOR patch that should also become part of this. This is a separate IOR than specular since skin can have multiple layers and so this IOR can be considered to be for another layer.

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.

20 Likes

Is there a chance that Principled V2 components like Multiscatter GGX land in 3.5 or are they planned for 3.6?

6 Likes

@brecht I guess the year in the title is incorrect. :slightly_smiling_face:

1 Like

Any Principled BSDF changes are for after 3.5.

Thanks, fixed.

4 Likes