CGNICO/UnrealEngine
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
c11d382a6f0acddb0bc1e95ab9bc5f8fc3570b55
UnrealEngine/Engine/Source/ThirdParty/OpenVDB/openvdb-12.0.0/tsc/meetings/2020-09-01.md
Jeffreytsai1004 c11d382a6f ue5-main
2025-05-18 13:04:45 +08:00

4.5 KiB
Raw Blame History

Minutes from 63rd OpenVDB TSC meeting, Sep 1st, 2020, (EDT)

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

Additional Attendees: Johannes Meng (Intel), JT Nelson (Blender), Andre Pradhana (DW), Ahmed Mahmoud

Regrets: Peter C.

Agenda:

  1. Confirm quorum

  2. Secretary

  3. Forum

  4. 7.2

  5. NanoVDB

  6. GCC 9.1 (PR 769)

  7. CpuTimer using C++11 chrono (PR 690)

  8. Introducing a "feature/abi8" branch (related to PR 788)

  9. Deprecating code (PR 806, StringGrid, Tree::prune(), LeafNode::str(), etc)

  10. "Locking down the grid configuration"

  11. Extrapolate SOP

  12. Next Meeting

  13. Quorum was confirmed.

  14. Secretary was Dan Bailey

  15. Forum

Ken to reply to post about large VDBs/OOC. Dan to reply to delayed-loading post.

There was a question about static libraries on Windows. Include paths need to be included as a system header using MSVC. Including VDB headers is generating warnings that are suppressed when building the core library. Nick has replied to this qs.

  1. 7.2

Nick has merged the PRs that removed the Makefiles and restructured the codebase. Planning on introducing the first PR that brings AX into the master branch, initially the core AX library only, then the Houdini SOP subsequently.

  1. NanoVDB

NanoVDB probably aiming for an 8.0 integration to allow a little more time for the codebase to mature. Ken currently merging NanoVDB up to the feature branch on a weekly cadence.

New work includes a tool for recomputing the grid statistics, foreach, range and invoke wrappers around TBB functionality. Current effort is in adding DirectX support, seems this will be relatively trivial.

Some users are asking for documentation around the memory layout. Investigation ongoing to look at replacing Jeff's work with the C library with a more automated approach. Jeff notes that initial compile time is a bit of a problem when using JIT.

Matt Pharr has incorporated NanoVDB into PBRT with pleasing results.

  1. GCC 9.1 (PR 769)

Addressed concerns Jeff raised with bool specialization for higher-order intepolation schemes by tackling the problem at the source. PR is now ready to be merged.

  1. CpuTimer using C++11 chrono (PR 690)

Dan investigated thread-safety issues with C++11 chrono. Adopted similar solution to that favored by TBB which is to store the number of microseconds since epoch instead of the time_point struct. This allows for starting and stopping a CpuTimer across multiple threads. PR is now ready to be merged.

  1. Introducing a "feature/abi8" branch (related to PR 788)

Need somewhere to put ABI=8 development in the run-up to 8.0.0. Solution preferred by all is to use the master branch instead of introducing a new feature branch that needs to be kept in sync. This will be gated by a standard ABI=8 macro but with an additional EXPERIMENTAL flag to reduce chance it will be picked up by accident, similar to the DEPRECATED flag.

  1. Deprecating code (PR 806, StringGrid, Tree::prune(), LeafNode::str(), etc)

When should we deprecate? Is one minor version sufficient? Consensus that we should be a bit more aggressive with deprecations, users often ignore deprecation warnings anyway. Still nice to give users as much warning as possible. New major version is a good time to drop API support and force users to change their code. We should consider introducing a 7.1.1 patch version with deprecation warnings. All to review the CSG operation PR 785.

Can we drop support for duplicateSourceStealable()? Nick still using it heavily. Limitations with current SOP implementation that causes issues with time dependency. Solution is to migrate to DS, but big sweep required to do so. Agreement that we should remove these methods anyway, Jeff/SideFX keen for people to not use this technique now.

  1. "Locking down the grid configuration"

Dan has implemented explicit template instantiation for the tree hierarchy. It speeds up building unit tests by around 30%, but building the core library now slower. It works by suppressing implicit template instantiation whenever tree hierarchy headers are included and explicitly instantiating them once in a source file compiled with the core library. Unfortunately had to be implemented using C preprocessor. More discussion needed, implementation shared as draft PR 813.

  1. Extrapolate SOP

Andre reports an issue with an assert firing in the fast sweeping code when run in debug mode, despite it working correctly in release mode. Assert in question validates that array indices are non-negative. Andre to share an example with Dan to investigate the root cause.

  1. Next Meeting

Skipping next week. September 15th, 2020. 1pm-2pm EDT (GMT-4).