Academy Software Foundation Technical Advisory Council (TAC) Meeting - February 4, 2025

Join the meeting at https://zoom-lfx.platform.linuxfoundation.org/meeting/97880950229?password=81d2940e-c055-43b9-9b5a-6cd7d7090feb

Voting Representative Attendees

Premier Member Representatives

Project Representatives

Carol Payne - D&I WG / OpenColorIO Representative
Cary Phillips - OpenEXR Representative
Chris Kulla - Open Shading Language Representative
Diego Tavares Da Silva - OpenCue Representative
Jonathan Stone - MaterialX Representative
Ken Museth - OpenVDB Representative

Industry Representatives

Jean-Francois Panisset - Visual Effects Society

Non-Voting Attendees

Non-Voting Project and Working Group Representatives

LF Staff

David Morin - Academy Software Foundation
Emily Olin - Academy Software Foundation
John Mertic - The Linux Foundation
Yarille Ortiz - The Linux Foundation

Other Attendees

Bill Ballew - Dreamworks
Doug Walker - Autodesk / OCIO
JT Nelson - Pasadena Open Source consortium / SoCal Blender group
Andrew Pearce - Dreamworks
Josh Bainbridge - Framestore
Lee Kerley - Apple
Lorna Dumba - Framestore
Olga Avramenko - Sony Imageworks
Rob Rowe - Cinepaint
Spencer Stephens - Zero Trust WG

Antitrust Policy Notice

Linux Foundation meetings involve participation by industry competitors, and it is the intention of the Linux Foundation to conduct all of its activities in accordance with applicable antitrust and competition laws. It is therefore extremely important that attendees adhere to meeting agendas, and be aware of, and not participate in, any activities that are prohibited under applicable US state, federal or foreign antitrust and competition laws.

Examples of types of actions that are prohibited at Linux Foundation meetings and in connection with Linux Foundation activities are described in the Linux Foundation Antitrust Policy available at linuxfoundation.org/antitrust-policy. If you have questions about these matters, please contact your company counsel, or if you are a member of the Linux Foundation, feel free to contact Andrew Updegrove of the firm of Gesmer Updegrove LLP, which provides legal counsel to the Linux Foundation.

Agenda

General Updates
- Security Threat model analysis for ASWF projects #615
- OpenQMC #434
rawtoaces annual review #475
MaterialX annual review #486
OpenVDB annual review

Notes

General Updates
- Security Threat model analysis for ASWF projects #615
  - Cary (vis Slack): OpenEXR met with the team a couple weeks ago for a kickoff but haven’t heard anything further from them
- OpenQMC #434
- Open Source Forum: starts at 12 for TAC members next week, make sure to be there a bit before. Larry: joint TAC / Governing Board meeting, do any of the members think there are specific TAC / project related topics that need to be included in the briefing? Let John / Carol / Larry know, if there is news you want to air to a larger audience. John: last such larger meeting was at SIGGRAPH in Denver, there will be significant senior leadership present. David: if you want to communicate to larger forum meeting at 2PM, I can plug in project news in the introduction. David: highly encourage presence / participation. John: opened issue for feedback in next meeting.
MaterialX annual review #486 - Jonathan
- Slide Deck
- Material X Project
  - Represent CG materials independently of individual tools and renderers
  - Launched at ILM in 2012
  - First used on Star Wars - Force Awakens in 2015
  - Released in 2017
  - Joined ASWF in 2021
  - Graduated as adopted project in 2024
- Active Contributors 2024
  - Using LF Insights chart
  - Largest uptick in September through Dev Days
  - January also a big month
- Contributing Orgs in 2024
  - ASWF -> mostly Jonathan / leadership
  - new contributions from Apple, NVIDIA, also from ILM
  - Additional contributor list
- Contributing Regions 2024
  - 36% US
  - 21% Canada
  - 17% Germany
  - 7% UK
  - 5% New Zealand (Weta)
- Third-Party Integrations
  - New in 2024: Octane Render in 2026 alpha, Substance Viewer from Adobe, still in beta
  - Houdini / Unreal Engine / V-Ray / Visio Pro / USD / Arnold / NVIDIA Omniverse / three.js / Maya / Max / …
- Highlights : MaterialX 1.39.2 Release
  - New Features
    - Chiang Hair BSDF (NVIDIA)
    - Generic Color Ramps (Autodesk)
    - Improved Worley Noise (SideFX)
    - Disney Principled Shading graph
    - Shader generation optimizations
  - Dev Days 2024
    - 10 merged PRs
- OpenPBR 1.0 Release
  - Long time collaboration between MaterialX, Adobe, Autodesk
    - With additional contributions from NVIDIA, Blender…
- Highlights: AOUSD Materials WG
  - Alliance for OpenUSD launched Materials WG in 2024
  - Developing a normative specification for OpenUSD materials
  - The MAterialX spec will be included via reference
  - This requires sections of the MaterialX spec to become normative
- Roadmap: Render Validation
  - A strong need for accurate comparisons of complex materials across renderers and languages
  - Initially focused on MaterialX< but will extend to OpenUSD assets in the future
  - Collaborations with the Open Shading Language team and Alliance for OpenUSD will be hugely valuable
- Roadmap: Open Social Platforms
  - MaterialX has recently migrated to BlueSky and Mastodon for public announcements and discussions
  - We’d love to build an ASWF community on these open platforms
  - Let us know if you are interested in joining and connecting with us
- TAC Open Discussion
  - Larry: can you bring up any negatives, what are you having trouble with, is there anything the TAC / ASWF can do to help? Jonathan: no major negatives, but the slides at the end is where we would like to work more with other groups, for instance we can’t do render validation on our own. Also we can’t migrate platforms on our own, we’d love to have a discussion about other groups, and see if we can come to a consensus. Emily: looking to set up BlueSky account for ASWF before Open Source Forum. Mastodon will need more work. We don’t have bandwidth to engage with community, right now it’s more used to just push information. How much community engagement can we do effectively? Wave: advantage of Mastodon is that if we put out an announcement, those of us on that can “boost” that. Jonathan: MaterialX is on both platforms, not sure which will “win”, but we’re happy to boost ASWF posts on both platforms. John: sounds like Emily is already on that. OCIO has a WG within their project, so if that’s something that works for MaterialX, we can help / support. Jonathan: already collaborating with OCIO on Nanocolor. Collabs are happening, wanted to highlight how valuable they are. John: every ASWF contribution in LF Insights came from Jonathan, it’s an attribution thing. Once data refreshes his affiliation will reflect. Jonathan: charts give me info that would be hard to gather myself. John: we’ll bring the LFX Insights team back here in a couple of months.
OpenVDB Annual Review - Ken
- Sparse Volume Tools build on VDB Data Toolbox
- Large Toolbox
  - Ever growing
  - Construct volumes and manipulate them
- List of 3rd party integrations
  - Newcomers are DCCs working with 3D Printing and topology optimization
- Release in 2024
  - Version 12.0.0 October 2024
  - Mozilla v2.0 -> Apache v2.0 license (requested by Apple)
  - A tedious process, people on the TAC were very helpful
  - Dropped support for GCC9
  - No Boost dependency if delayed-loading is disabled
    - Hope to completely remove it as some point, but still need it on Windows for delayed loading
  - New API methods for low-level tree manipulations
  - Major cleanup of NanoVDB to mimic OpenVDB (namespace etc)
  - Python bindings have been ported from PyBind11 to nanobind
    - has better support for GPUs
  - IO streaming of raw NanoVDB buffers (no headers needed)
    - Serialization of trees exist in a single memory block
    - No overhead from passing headers
- OpenVDB wins Sci-Tech Award
- Release in 2025
  - Two minor releases per year
  - One major release per year
  - Increase cadence
  - Version 12.x feb 2025
    - SDFs from tapered tubes (think 3D wire-frames)
    - In-memory half support (requested by Autodesk)
    - Anisotropic surfacing (elliptic particle footprints)
    - Support fo latest CLANG (requested by Apple)
    - Multi-GPU tools in NanoVDB and fVDB (new project)
- fVDB : Spatial-Learning at Scale
  - research project I lead at NVIDIA (11 people)
  - Paper at SIGGRAPH
  - Neural net architectures come down to a few fundamental operators: CNN, attention
  - Many existing open source implementations: TensorFlow, PyTorch, Jax, limited to 2D data (images) or 1D data (strings), assume data is dense
  - Want to use operators on the real world, 3 dimensional, very large
  - Neural Radiance Fields at very large scale, gaussian splats
  - Reconstructing the world from scanned data
  - Deep learning on volumetric data is hard
    - 3D data is
      - Sparse
      - non-uniform
      - Too big to fit in RAM
  - Want to developer a unified framework for end-to-end deep learning on large 3D Data
    - Small academic libraries exist that have some operators.
  - fVDB
    - Sparse representation for the features
    - fVDB is a unified framework for deep-learning on large-scale, high-resolution volumetric data built on top of PyTorch
  - The VDBs : a quick attempt to reduce ambiguity
    - VDB is the data structure (2009)
    - OpenVDB is open source release of VDB (2012)
    - NanoVDB: GPU friendly version of VDB (2020)
    - NeuralVDB: neural compression of VDB (paper in 2023), a neural compression technique on top of VDB trees.
    - fVDB: framework combining… (2024)
  - Unified API for
    - Building and training neural networks…
  - Feature Comparison
    - Runs on top of CUDA
    - Memory efficiency
    - multi-GPU support
  - Core Building Blocks
    - JaggedTensor: handling batches of non-uniform data efficiently on the GPU
    - GridBatch
    - Accelerated Operators
    - MultiGPU
  - NanoVDB change to work on neural data: in OpenVDB everything is “mashed together”, data and ijk coordinates. Once we separate data from structure, can use for machine learning.
  - Surface Reconstruction
  - World-scale NeRFs
  - 3D Generative AI
  - Built on top of NanoVDB
- Blocking Issue
  - fVDB lives in feature branch
  - Need to resolve CI
  - Amzon Cloudbuild access is needed to fVDBs CI test to use GPUs
  - This is blocking merge of fVDB into OpenVDB main branch
  - Made a request in October but we’re still waiting on LF support to insert credential keys on Github repo (Jonathan Schwartz)
    - John: escalating internally, will follow up.
- Comments?
  - Larry: exciting work
  - John: do we still have quorum? JF: yes
  - Erik R: can you talk more about relicensing to Apache 2? In OpenAPV, currently licensed as BSD, curious if that will present similar challenge. Ken: I don’t know why Apple objected to Mozilla 2 license, don’t recall getting a “good” answer. Had something to do with clause on patents? But BSD should be less of an issue. Was extremely painful, everyone who every committed anything to approve, was not easy. Michael Johnson: one license was very straightforward to get through legal to allow commits, the other “not so much”. Ken: no push back once we got project running, just took a long time. A good change in the end. John: if another project wants to do this, reach out to us, we can help guide the project through the process. Larry: OIIO did a license switch from BSD to Apache2 when it joined the foundation, wasn’t too troublesome, everybody we could contact (except one) did it immediately. One person was upset they were contacted at all, an old contribution. Easier for projects where the majority of the code was written by a small group of people you can reach. We had a long tail, but we easily got to 90%. A couple of projects have done it, it is possible, hit us up if you want to do it.
  - Larry: can you share wisdom with switch to nanobind? Ken: I don’t think I can comment much, it was an ask from inside NVIDIA, change was relatively painless, builds are faster and smaller. Can ask the person who pushed for that to provide more details. But can recommend it, no drawbacks so far. I invited him to the TAC a few months ago. He has slides, I can forward them. Larry: was wondering if experience was overall positive. There was discussion in other TSCs about doing this. There can be friction for projects working together that don’t match. Ken: a winning argument was reading PyBind11 docs which recommend using nanobind. Larry: came up in OpenEXR, author recommended switching as well.
Motion to review both projects?
- Both projects are renewed