900: Fix initializing buffer with unaligned data r=grovesNL a=Uriopass
In wgpu 0.7, this was a valid operation, the zero padding was done by hand.
In #872 this logic was removed as an optimization since buffers are already zero initialized, however the slice end was taken from buffer.slice which has to be aligned when mapped.
This is causing issues when trying to update imgui-wgpu-rs (https://github.com/Yatekii/imgui-wgpu-rs/issues/54) to wgpu 0.8 since they use U16 index buffers with odd lengths.
Co-authored-by: Pâris DOUADY <paris.douady@hotmail.fr>
In wgpu 0.7, this was a valid operation, the zero padding was done by hand.
However in #872 this was removed as an optimization since buffers are zero padded, however the slice end was taken from buffer.slice which has to be aligned when mapped.
894: Skip mapping of zero sized buffers in create_buffer_init to avoid a panic and avoid creating an unused padded buffer r=kvark a=Imberflur
Co-authored-by: Imbris <imbrisf@gmail.com>
* Add validation-windows yml for HLSL
Add workgroup size to hlsl-out
* Update Readme with Make target information
* Change runs on to windows-latest
* Remove snapshot file
* Add support for arrayLength to the wgsl frontend
* Fix clippy warning
* Add draft support for array length to the msl backend
* Finalize support for array length in msl
* Convert buffer size to array length in msl backend
* Fix clippy warning
* Fix misleading documentation
* Changes based on review of PR
* Use a fake binding for sizes buffer in msl backend if necessary
* Only generate the msl buffer size structure if globals are present that have unsized arrays
* Make sure to generate separators
* Replace uses_unsized_buffers with !self.runtime_sized_buffers.is_empty()
* Clear self.runtime_sized_buffers
* Run snapshot of shadow.msl
* Add Expression::ArrayLength support to the spirv backend
* Remove wgsl target from the access snapshot test
* Update access.msl snapshot
* Update stack size for msl backend
* Update stack size again for msl backend