3.9 KiB
Minutes from 62nd OpenVDB TSC meeting, Aug 25th, 2020, (EDT)
Attendees: Nick A., Jeff L., Ken M., Dan B.
Additional Attendees: Johannes Meng (Intel), JT Nelson (Blender), Bruce Chernia (Intel), Andre Pradhana (DW), Ahmed Mahmoud
Regrets: Peter C.
Agenda:
-
Confirm quorum
-
Secretary
-
Forum
-
Siggraph Recap
-
Roadmap
-
Next Meeting
-
Quorum was confirmed.
-
Secretary was Nick Avramoussis
-
Forum
Ken to reply to post about large VDBs/OOC.
- Siggraph Recap
Good response from the ASWF Open Source days, presentations are now live. Good idea to present the roadmap. Multiple questions w.r.t NanoVDB vs GVDB, it would be good to clarify to the community the differences and scopes. Ken, comparing the two directly is difficult.
ASWF forum vs Google forum question - unclear reply. We should dedicate a meeting soon to come up with a stance. Dan is monitoring the ASWF slack. We could deprecate the google forum. Ken, would be a shame to lose users. TBD.
- Roadmap
We only have a few months for a 7.2/8 release window. Ken, main focus is NanoVDB, Nick, main focus is AX. AX is ready to go. OpenVDB repo needs some adjustments before a PR can be made. Ken, feature branch approvals is frustrating, Dan to investigate if it can be removed.
Dependencies are a big focus. Ken, NanoVDB uses a utility header/wrapper around TBB to abstract away the implementation. Dan, all we need is a threaded/ non-threaded switch to begin with and users can subsequently implement their own threading constructs. Nick, TLS constructs are slightly tricker. The wrapper should all be task based threading which provides an interface for others to extend. All agree this is a worthwhile investment to port to OpenVDB. Dan to investigate some of the other dependencies - primarily zlib, half and boost (w.r.t delay loading). Need to make sure not to break ABI with any delay loading changes. However ABI 8 could provide a nice opportunity for more substantial change. All skeptical of how useful the per leaf delay loading granularity thats currently implemented is - it convolutes the tree. Lots of work in this area, no obvious/agreed on solution, though enabling/disabling through an optional dependency is a good start.
Makefiles, vote to remove. Unanimous consent. Nick to action. Bump to CMake 3.12. Unanimous consent. Nick to action. Look over deprecated code for removal for 7.2/8.
Pruning of tree methods. Dan, proposed new CSG methods make a bunch of methods (primarily the visit methods) obsolete and could be deprecated. All agreed that this concept is still worth pursuing. Dan, once this PR is in, can contribute the Merge SOP.
Locking down the tree structure. Non-obvious timeframe, would most likely be significant work. Jeff, for VDB 8, see if we can make it harder for this to be customized or to build in compiler warnings for users who are doing so.
Tile support for all tools. Ken, there's no one size fits all - each tool will treat tiles differently. Nick, will push an extension to Filter.h which supports tiles. Ideally this should be generalised such that all tools can call the same tool to configure tile topologies. Indirect goal to remove the Densify SOP.
Andre, continuing to work on the Extrapolate SOP for 7.2/8. Seems to be a problem with extensions of non-float grid types. Andre to provide an example to Ken to fix.
Ken to take another look at the sharpen PR #756 and merge. Nick, Morphololgy PR #754 is ready for review and can be targeted for 7.2.
JT, preparing a demonstration of VDB in blender, both on master and other more experimental features.
Dan, will merge CI changes.
Dan, release notes conflicts are still annoying. Jeff, can we implement custom conflict rules for github. All agreed this would be ideal, but unsure on githubs support for this. Dan to investigate.
In future meetings, we should discuss how to better organise the roadmap for "good first issue" tasks to on-board more developers.
- Next Meeting
September 1st, 2020. 1pm-2pm EDT (GMT-4).