Minutes from 51st OpenVDB TSC meeting, June 4th, 2020, (GMT)

Attendees: *Nick* A., *Jeff* L., *Ken* M., *Dan* B.

Additional Attendees: Johannes Meng (Intel), JT Nelson (Blender),
Peter Cheng (DW), Andre Pradhana (DW), Bruce Chernia (Intel),
Ibrahim Sani Kache (DW)

Regrets: *Peter* C.

Agenda:

1) Confirm quorum
2) Secretary
3) Fast Sweeping
4) Python 3
5) Other
6) Next meeting


1) Quorum was confirmed.

2) Secretary was Nick Avramoussis

3) Fast Sweeping

Ken has pushed up his implementation of his approach to the re-computation of
signed distance fields, Fast Sweeping:
http://www.museth.org/Ken/Publications_files/Museth_SIG17.pdf
The code is available on Ken's fork of OpenVDB:
https://github.com/kmuseth/openvdb/tree/OVDB-133_FastSweeping
Ken to open a draft pull request into VDB for others to comment on and trial the
implementation. Ken, some work still to be done however the bulk of the
algorithm is complete. There exists a variety of new public methods to perform
re-computations of level sets, conversions of scalar fields (signed fog) to
level sets and the value extrapolation of level sets, potentially coupled with
another scalar grid. CI build is currently failing due to some strict C++
checks which should be easy to fix. It was noted that the extension of anything
other than a scalar grid (i.e. vector grid) currently requires multiple
invocations of the algorithm, something that could be improved.

4) Python 3

Ibrahim asked about python 3 support and OpenVDB. We currently have CI images
that run in VFX 2020 which test with 2020's python version, currently at 3.7.
There have been some reported issues with python 3.8, however this should only
be specific to the architecture of the unit tests and not actually the python
plugin itself. TSC unaware of any other issues regarding python 3. Note that the
VFX 2021 container has now be drafted.
https://vfxplatform.com/

5) Other

Discussion into OpenVDB's dependencies. Ken, one of the biggest complaints he
hears is VDB's dependency list. Having some of these dependencies as VFX
platform candidates could help - mainly referring to blosc. Jeff, LLVM (as a
requirement of AX) should not really be a part of the VFX platform (due to its
volatility w.r.t API/ABI), with downstream software instead opting for always
hiding LLVM symbols. Boost python is heavily relied on - pybind would be a
reasonable alternative but that also does not appear as part of the VFX platform.
Note, OpenEXR have decided to split out Imath and potentially advocate it as a
math library for VFX:
https://github.com/AcademySoftwareFoundation/Imath

Ken mentioned open source licensing rules for contributors - would it matter if
an addition to OpenVDB was made under a different Open Source license? Nick, AX
has opted to use the same license as OpenVDB so there are no concerns there.
Potentially would have to be deferred to the Linux foundation.

6) Next meeting

June 18th, 2020. 2pm-3pm EDT (GMT-4).