523: Change buffer binding to not take a slice r=cwfitzgerald,grovesNL a=kvark
This has been a constant source of confusion w.r.t dynamic offsets.
What it looks like it's doing: user provides a slice of buffer visible to the shader. Just like they provide slices to vertex and index buffer bindings.
What it's actually doing: it takes the offset and size separately, modifying the offset later on based on the dynamic offset.
We don't need this confusion. Semantics is different from `BufferSlice` and so we should have a different API here.
Co-authored-by: Dzmitry Malyshau <kvarkus@gmail.com>
517: Added some explanatory comments to hello-compute example r=kvark a=JonathanWoollett-Light
Simply added some more comments explaing the basic compute example.
Also moved the lines:
```rust
let slice_size = numbers.len() * std::mem::size_of::<u32>();
let size = slice_size as wgpu::BufferAddress;
```
To be just before where they are used.
Co-authored-by: Jonathan Woollett-Light <jonathanwoollettlight@gmai.com>
509: Convert StagingBelt to take a &mut CommandEncoder r=kvark a=cwfitzgerald
This allows `StagingBelt` to be `Send`, very important for async/highly multithreaded applications. Also added a bit more documentation.
Tested on the skybox example.
Co-authored-by: Connor Fitzgerald <connorwadefitzgerald@gmail.com>
502: Expose get_bind_group_layout r=grovesNL a=kvark
Returning a new object is fine for BGLs specifically and for these cases in particular because we made sure the refcount is bumped (and BGL has a special kind of refcount) in f164fd1f46
Co-authored-by: Dzmitry Malyshau <kvarkus@gmail.com>
495: Update a documentation to refer to create_buffer_init() instead of create_buffer_with_data() r=kvark a=yutannihilation
A followup of #482
Co-authored-by: Hiroaki Yutani <yutani.ini@gmail.com>
493: Use BCn textures if possible in skybox example r=kvark a=cwfitzgerald
This converts the skybox to using compressed textures if available. I have designed the code to be extensible to other compressed formats as they are added.
Additionally I added srgb to the texture types, as we were using a srgb framebuffer without properly converting into linear on the texture reads.
Co-authored-by: Connor Fitzgerald <connorwadefitzgerald@gmail.com>
492: Replace NonZeroU32 with u32 in TextureViewDescriptor r=kvark a=kunalmohan
Includes https://github.com/gfx-rs/wgpu/pull/854
(rustfmt seems to have introduced a number of other changes. I can revert them if they are undesired)
r?@kvark
Co-authored-by: Kunal Mohan <kunalmohan99@gmail.com>
482: Add label parameter to `create_buffer_with_data` r=kvark a=OptimisticPeach
Improves QOL by allowing you to attach a label to a buffer when created with predetermined data:
```
let uniform_buf = device.create_buffer_with_data(
bytemuck::cast_slice(my_data),
wgpu::BufferUsage::UNIFORM,
Some(Cow::Borrowed("My Uniforms")),
);
```
Co-authored-by: OptimisticPeach <patrikbuhring@yahoo.com>
484: Enable features in CI to prevent compiler errors r=kvark a=cwfitzgerald
Simple change, this also builds features in CI.
I added trace and replay features to wasm, even though wasm doesn't support tracing as is, because people may enable them to get serde support for wgpu-types stuff.
Co-authored-by: Connor Fitzgerald <connorwadefitzgerald@gmail.com>
