You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This issue will track short term work that needs to be done so people can quickly pick up tasks.
High priority
Fix HL
Clean up code style. C++ native code should match ozz code style, haxe-side functions should use haxe styling.
General
Model definitions are currently based off the sample. need to modify these to match Heaps more closely (255 bone limit, etc) and remove the "ozz sample" tagging. Then test importers and import new test model for sample
Need to check the FBX importer builds properly. This involves grabbing the FBX SDK and making sure the cmake file detects it properly and does all the right things. Should work?
Can we use github actions to automatically generate hdll and wasm files?
Native Extension
Memory ownership model needs to be looked at. Currently, skeletons and meshes can't be shared between models, resulting in significant amounts of over allocation and excessive reading. There may be other instances where memory ownership is bad.
Fix HL padding. There's probably a less dumb way to do what I'm doing, and it's allocating too much right now anyway.
Skin matrices get copied multiple times per frame, we should reduce this as much as possible. Ideally we would be able to copy straight to the shader and upload from there; this MIGHT be possible with shared memory?
Mesh and Model currently hang on to some buffers internally. Is this the right place to do this? Should they be moved to be owned by haxe?
Heaps integration
Currently meshes render all parts as one contiguous block. need to break this into parts for multi material to work I think?
Need to create fake joint objects so you can track and attach to bones at runtime. This probably involves pulling data out of the skinning matrices.
Sample heaps integration currently is a bare bones impl. Need a proper integration with heaps resource system (support .toModel(), etc)
Current model/mesh/prim layout probably needs cleaned up. Ownership model is weird, probably not very heaps-like. Look at how Library works with hmd models.
Fix it so we can actually cache and share model objects between instances.
Toolchain
Influences count needs to be fixed at 4 on the importer to match the skinning shader. Currently you can specify a maximum, but not a minimum.
Need to figure out a reasonable way to integrate with heaps resource system. Currently the tools will spit out separate files from a model, and require specifying a skeleton to export a mesh. FBX files should auto-generate _mesh, _skeleton, and animation exports for files with geometry, and just animation for files that only contain animation (No need to re-export these skeletons)
The text was updated successfully, but these errors were encountered:
This issue will track short term work that needs to be done so people can quickly pick up tasks.
High priority
General
Native Extension
Heaps integration
.toModel()
, etc)Library
works with hmd models.Toolchain
The text was updated successfully, but these errors were encountered: