2.7 KiB
Minutes from 145th OpenVDB TSC meeting, September 6th, 2022, (EDT)
Attendees: Ken M., Dan B., Rich J. Andre P.
Regrets: Nick A., Jeff L.
Additional Attendees: Peter Cheng (DW), Karl Marett, Sebastian Gaida, JT Nelson (Blender), Greg Hurst (United Therapeutics), Bruce Cherniak (Intel)
Agenda:
- Confirm quorum
- Secretary
- Forum and Discussion a) Request to expose 4x4 matrix in Python bindings b) Update build instruction
- Failure in NanoVDB CI
- VDB in medical imaging (OpenVDB Slack Channel)
- Mathematica Link Update
- Update from DreamWorks
- CI Effort
- Root Node Offset
- Next Meeting
- Confirm quorum
Quorum is present.
- Secretary
Secretary is Andre Pradhana
-
Forum and Discussion a) Request to expose 4x4 matrix in Python bindings. This should be done through PyBind 11. A few people expressed an interest to help with this task.
b) Update build instruction to reflect the correct version of TBB.
-
Failure in NanoVDB CI
We look at failure NanoVDB CI failure in MacOS. Ken has an idea that it may be caused by the new allocation that is implemented.
- VDB in medical imaging (OpenVDB Slack Channel)
Somebody was asking about the potential of using OpenVDB or NanoVDB for medical imaging purposes. The answer is: yes, VDB can be used for this purpose.
- Mathematica Link Update
Greg is still waiting for a reply from Wolfram.
- Update from DreamWorks
DreamWorks said that they will go through with a more thorough investigation of existing old VDB files that uses the dense table in the root node. They will probably convert old VDB files to new VDBs.
- CI Effort
Dan gives an update on the CI effort. We have an access to two machines provided by AWS.
Dan claims that the billable hours for all CI to finish is 1 hr and 43 minutes.
- Root Node Offset
Ken revisited his work on the root node offset. He did not find any performance issues with the new implementation. Several things to note: a) The way that Ken solves the problem is by adding a CoordT origin in the root node to offset the position of the children of the root node. b) The offset should coincide with the grand-child of the root. In a regular tree configuration the offset should be 2048 = 16 * 16 * 16 / 2. c) Ken proposed the origin offset on the root-node to be dynamic, because in the future, we can read an old VDB file with the root node origin set to be (0, 0, 0).
He proposed that we would go through with the ABI change, i.e. adding the origin member data in the root. He also proposed that in the future, doing operations on two or more different grids should be restricted on grids with the same origin.
- Next meeting
Next meeting is on September 13th, 2022. 1pm-2pm EDT (GMT-4).