Minutes from 99th OpenVDB TSC meeting, July 20th, 2021, (EDT)

Attendees: *Ken* M., *Jeff* L., *Andre* P, *Nick* A.

Additional Attendees: JT Nelson (Blender), Bruce Cherniak (Intel),
Johannes Meng (Intel) , Jeff Budsberg (DW) , Richard Jones (DNeg)


Regrets: *Dan* B.

Agenda:

1) Confirm Quorum
2) Secretary
3) Google Forum
4) NanoVDB
5) 9.0
6) String Grids
7) Example Hip File
8) Next meeting


1) Confirm Quorum

Quorum is present.

2) Secretary

Secretary is Jeff Lait.

3) Nano VDB

We had agreed to split the viewer to a new branch.  One branch will be continuation of existing one, the other is a merge candidate.  Sync to 8.1 has been complete.  Removing the viewer has proved more difficult than expected.

4) VDB 9 / VFX Platform

We plan on releasing 9 around Siggraph.  VFX platform will not put 9 in there until we are out.  We end up forced into pushing the EXR support early.  

5) EXR

Patch has been submitted.  This suggests the maintainers want the renderer to still build.

6) 9 requirements

Do we put NanoVDB in it?  We should aim for this, but it doesn't affect VFX compatibility.

We need to have any ABI changes.  

Root Node Offset.  The file format doesn't have to change.  The base root node is the only one that gets moved, all the internal nodes have the same topology.  Ken will investigate this.  We can also change our root search method at the same time.

Blind data.

Other ABI 9 pending the flag changing to be done.

Last sweep to make sure there isn't anything to take out of leaf nodes to make them more light weight.

7) PR 1110 Deprecating String Grids

What are our requirements for types?  Do we support anything other than POD?

It is very inefficient so why support that?  It is unclear when destructors are called in our container.

Proposal: VDB will only support "simple" datatypes without dynamic memory management.  We will discuss again when Dan is present.

If we eliminate the string grid, but don't prevent others from implementing it, the PR is not a problem.   We could leave unit tests with a manual specification.  If we find something that would allow us to accelerate by assuming a simpler type we could address it then.  Note that serialization isn't working in any case.

We can deprecate the node union specialization.  This is currently using the C++ standards to detect it is safe.  Depending on 8.2 release date, we could remove for 9.

8) PR 1106

One-sided extrapolation.  Should be reviewed.

9) PR 1091

Expansion pattern, approved and should go in.

10) PR 1097

Fix for size_t that generalizes the macOS solution.  Needs one more approval.

11) Open Source Day

Discuss the Open Source day next meeting.

12) Next meeting

Next meeting is July 27th, 2021. 12pm-1pm EST (GMT-5).
It is apparently our Hundredth Meeting!