Merge pull request #2004 from sideeffects/sendupsream_tscnotes20250205

Meeting notes for Feburary 5th, 2025.

Author: jmlait

tsc/meetings/2025-02-05.md

@@ -0,0 +1,92 @@
+Minutes from OpenVDB TSC meeting, Feburary 5th, 2025
+
+Attendees: *Ken* M., *Andre* P, *Dan* B.,  *Jeff* L., *Nick* A., *Greg* H.
+
+Additional Attendees: None
+
+Regrets: *Richard* J.
+
+Agenda:
+
+1) Confirm quorum
+2) Secretary
+3) OpenVDB Course
+4) Value Accessors
+5) Half-Float Support, PR2002
+6) PR1995
+7) PR1919
+8) Next meeting
+
+------------
+
+1) Confirm quorum
+
+No quorum.
+
+2) Secretary
+
+Secretary is Jeff Lait.
+
+3) OpenVDB Course
+
+a) Rendering.  While used in rendering, nothing particularly novel or
+   interesting beyond "Yes, we used it."
+b) Permissions
+   i) We want the content to be shippable.
+   ii) Can we get the protected material excluded, or if not possible,
+       switch to a shorter ten minute taster that doesn't have
+       forbidden content?
+c) Order
+   i) Intro (Ken)
+   ii) Applications
+      1) Production Examples
+      2) MPM (Alexandre)
+      3) Capsules, United Therapeutics (Greg)
+   iii) fVDB
+   iv) Coding
+      1) 16bit
+      2) Iterators, Parallel Patterns
+      3) Anistropic Surfacing
+         a) Theory?
+         b) Engineering of performance?
+
+Will submit an inclusive application.
+
+4) Value Accessors
+
+Little re-use on GPU?  You can use a complete access instead often.
+Add 8^3 hash table at root, so usually that gives free access, for no
+linear search and direct access, excpet for rare collisions.
+
+On CPU main bottle neck is getting an adjacent node, local -> global
+-> hash is the cost, vs directly accessing leaf node.
+
+Better native hardware support?  FMA?  v3 support?  Look into better
+use of intrinsics?
+
+Can use first static constructor to try to test instrumentation.
+
+5) Half-Float Support, PR2002
+
+Currently a DCO problem from commits that are already in the master
+branch.  Can do a force push to unify?
+
+Will try a new feature branch to see if that clears it up.
+
+6) PR1995
+
+Typo fixes.  Does change some actual code.   Would need to commit soon
+so it doesn't diverge.  But we really want to make sure nothing
+unintended slips through.
+
+Ask for it with no code, and have two poeple running over every line.
+
+7) PR1919
+
+Finishes removing CPPUnit test, will be examined by Andre.
+
+8) Next meeting
+
+Agenda: TBD
+
+Feburary 19th, 2025, usual time