Minutes from 89th OpenVDB TSC meeting, May 4th, 2021, (EDT) Attendees: *Ken* m., *Jeff* L., *Dan* B., *Andre* P. Additional Attendees: Jeff Budsberg (DWA), JT Nelson (Blender), Johannes Meng (Intel) Regrets: *Nick* A. Agenda: 1) Confirm Quorum 2) Secretary 3) Siggraph Course 4) ABI=9 4a) Root Node Offset 4b) Blind Data 5) NanoVDB 6) Example Hip File 7) Quick review of PRs 8) Next Meeting 1) Confirm Quorum Quorum is present. 2) Secretary Secretary is Dan Bailey. 3) Siggraph Course Siggraph course has been accepted with a very high review score. It will be a three hour session. Presenters will be Ken Museth, Jeff Lait, Jeff Budsberg, Richard Jones and Dan Bailey. There may be an additional presenter which would mean 30 minutes for each presentation. Jeff Budsberg will present the VDB Extrapolate SOP instead of Andre. Richard Jones will present VDB AX instead of Nick. Presentations must be pre-recorded and delivered on 4th June, or 11th June if delivered with closed captioning. All in agreement that it would be preferred for Siggraph to handle the closed captioning for us, so 4th June is the deadline for recordings. The videos will need editing and splicing together, JT mentions that he is a video editor so could help us with this. Course notes and supplemental material need to be delivered by 11th June. 4) ABI=9 What changes should be considered for ABI=9? New features can be added now and gated by the ABI=9 macro in preparation for the v9.0 release around Siggraph time. 4a) Root Node Offset A root node offset in space would be highly desirable as the common use case of objects centered on the origin could feasibly use just one root node child. This is expected to greatly accelerate random access, particularly when using the ValueAccessor as it would avoid needing to go through the root node hash table which is known to be slow. However, this is a theory and should be tested to confirm. Backwards compatibility is important, a breaking change to the file format is a big deal. One option is to apply the offset in-memory only during VDB deserialization. This would have the benefit of not requiring any changes to the file format itself, however the IO code will need to become more complicated as a result. Potential to test performance by manipulating a VDB to apply the offset and accessing the impact on the root node access times. Open question about whether an index-space offset would still need to be applied for every voxel access and what exactly the cost of that might be. More investigation and thought needed here. 4b) Blind Data PR1050 attempts to implement this feature. After a brief discussion, the name transientData() is settled upon for the blind data method. This makes it clear that the blind data is not serialized. All in agreement that a 64-bit unsigned integer allows room for expansion at a later date without needing an API change. The proposed technique of stealing bits from the origin coordinate to extend the range is not part of this PR. 5) NanoVDB NanoVDB will be shared with the group for review soon. Likely just the core library at first. One missing component is updating PNanoVDB (the C API). SideFX needs an updated CNanoVDB API to match the new data layout, this may end up being something that is maintained on their end. 6) Example Hip File Jeff Budsberg has been updating the example hip file that has been shared on the website removing some old nodes and adding some new examples. Currently missing any examples for VDB AX as this is not yet released at DWA. Plan is to release the current version and then to extend nearer to Siggraph with some VDB AX examples. 7) Quick review of PRs PR1045 - VDB Activate SOP world space parameter and question about expand/dilate behavior. Jeff mentions that there is precedence in Houdini to want two constraints to be enforced. Suggestion is to add a warning instead. PR1038 - Dan to implement VDB Resample SOP feedback from Jeff. PR1031 - Dan has now implemented VDB Merge feedback from Jeff. 8) Next Meeting Next meeting is May 11th, 2021. 12pm-1pm EST (GMT-5).