The default gltf for android's gltf-viewer is the buster drone,
which uses skinning.
We add an additional model (damaged helmet) as an alternative
default. This will be useful for debugging WebGPU, where skinnning is
not yet supported. This alternate path can only be invoke by starting
the app through adb:
`adb shell am start -n com.google.android.filament.gltf/.MainActivity --ez "use-static-model" true`
- Re-enable docs presbumit check (checks if there has been direct
edits of /docs).
- Fix script to download and install mdbook
- Update Materials.md.html and re-run the build script
(build/run.py).
* Fix staging buffer alignment
Note: we were not aligning texture copies properly, so this broke
certain use cases on the Pixel 4.
* Address PR comment
Convert optional type to a simple int type with a default value, for
the alignment arg for acquiring a staging buffer.
* VK: Simple staging bypass for free uniform buffers
When the uniform VulkanBuffer is currently not in
use and under UMA, just perform a direct memcpy to
GPU memory.
Add a flag to control disable this flow and always
use staging to update buffers.
* Addressing review comments
Use a constant instead of a custommization option
to known if we always use staging for updates or not.
---------
Co-authored-by: Serge Metral <sergemetral@google.com>
`unfiterable : true` indicates that the sampling of the texture
will not apply filtering.
This is mainly to satisfy the webgpu requirement for
bindGroupLayouts.
- Documentation has been updated for both `unfilterable` and
`multisample`.
- Internal materials have been updated where necessary (depth
samplers have to be marked as `unfilterable`.)
- webgpu code has been changed appropriately.
BUGS=420745987
- Added an explicit link flag for libxcb in backend/CMakeLists.txt
- Made a string func inline, otherwise we'd get duplicate
definition error during linking.
FIXES=419788369
When a program was created directly from a cached blob in
`ShaderCompilerService`, its associated token was not signaled as ready
in the THREAD_POOL mode. This oversight caused a deadlock at the
`token->wait()` call during program destruction.
This commit resolves the issue by skipping the token's readiness check
upon destruction if the program was created from the cache blob.
BUGS=[423221474]
Powei Feng
Andy Hovingh
Benjamin Doherty
Doris Wu
Sungun Park
Anish Goyal
Jeremy Nelson
rafadevaiandSerge Metral
Juan Caldas
utzcoz
Konrad Piascik
Matthew Hoffman
bridgewaterrobbie
Sungun Park
dependabot[bot]andPowei Feng