Adds a `BindingResource` variant for external textures. In core's
create_bind_group() implementation, allow binding either external
textures or texture views to `BindingType::ExternalTexture` layout
entries.
In either case, provide HAL with a `hal::ExternalTextureBinding`,
consisting of 3 `hal::TextureBinding`s and a `hal::BufferBinding`. In
the texture view case we use the device's default params buffer for
the buffer. When there are fewer than 3 planes we can simply repeat an
existing plane multiple times - the contents of the params buffer will
ensure the shader only accesses the correct number of planes anyway.
Track the view or external texture in `BindGroupStates` to ensure they
remain alive whilst required.
And finally, add the corresponding API to wgpu, with an implementation
for the wgpu-core backend.
* WIP
* Fix typo
* WIP: Implement structure of command_encoder_transition_resources
* WIP
* More work
* Clippy
* Fix web build
* Use new types for API, more docs
* Add very basic test
* Try to fix test cfg
* Fix merge
* Missed commit
* Use wgt types instead of hal types
* Implement `Clone` for `ShaderModule` (#6939)
* Move to dispatch trait, move more things to wgt
* Move existing code to use new wgt types
* Fixes
* Format import
* Format another file
* Fixes
* Make module private
* Fix imports
* Fix test imports
* Rexport types
* Fix imports
* Fix import
---------
Co-authored-by: Alphyr <47725341+a1phyr@users.noreply.github.com>
Co-authored-by: Connor Fitzgerald <connorwadefitzgerald@gmail.com>
* Initial Commit
* Remove now-redundant `format` import
* Update CHANGELOG.md
* Appropriately feature-gate `texture_format_serialize` test
* Remove `alloc` feature
Also fixed some documentation links and a Wasm `std` import
* Revert change to `Serialize` for `TextureFormat`
* Combine use statements
* Switch from `PathBuf` to `String`
* Consider environmental flags as unset on `no_std`
* Fix missing `format!`
* Add new CI tasks for `no_std` testing
* Comment out known failing CI matrix option
* Update all usage of `Dx12Compiler::DynamicDxc`
* Added comments to CI
* Update .github/workflows/ci.yml
* Update .github/workflows/ci.yml
* Update .github/workflows/ci.yml
* CI Touchups
---------
Co-authored-by: Connor Fitzgerald <connorwadefitzgerald@gmail.com>
This commit was authored by running the following Nushell script, using
the `nu` binary and the lovely `fastmod` tool:
```nushell
# Copy-pasted from the OP in [`gpuweb`#4838](https://github.com/gpuweb/gpuweb/pull/4838).
let renames_table = '
Type Old New Used in
dict GPUImageDataLayout GPUTexelCopyBufferLayout "writeTexture,
parent type of ↙"
dict GPUImageCopyBuffer "GPUTexelCopyBufferInfo
extends ↑" T2B, B2T
dict GPUImageCopyTexture GPUTexelCopyTextureInfo T2B, B2T, T2T, writeTexture
dict GPUImageCopyTextureTagged "GPUCopyExternalImageDestInfo
extends ↑" copyExternalImageToTexture
dict GPUImageCopyExternalImage GPUCopyExternalImageSourceInfo copyExternalImageToTexture
union GPUImageCopyExternalImageSource GPUCopyExternalImageSource member of ↖
'
let renames_table = $renames_table
| from tsv
| select 'Old ' 'New '
| rename old new
| update new { $in | lines | get 0 } # only the first line has the renamed symbol identifier
| update old { strip_gpu_prefix | str trim }
| update new { strip_gpu_prefix | str trim }
| sort-by old | reverse # Replace most specific symbol names first (some have the same "word segments" but with fewer segments)
def strip_gpu_prefix []: string -> string {
$in | str replace --regex '^GPU' ''
}
# Rename image APIs.
for entry in $renames_table {
fastmod --accept-all --fixed-strings $entry.old $entry.new --iglob '!CHANGELOG.md' --iglob "!xtask/src/vendor_web_sys.rs" --iglob '!wgpu/src/backend/webgpu/webgpu_sys/' --iglob '!deno_webgpu/' --iglob '!wgpu/src/backend/webgpu.rs'
}
cargo fmt
```
…and cleaning up `deno_webgpu/`'s Rust compilation errors.
PR #6150 suffered a much larger rebase "hell" than I anticipated. On
my Linux box I made this change, but lost it while force-pushing from
Windows (and created some other compiler errors while at it...).
By disabling all features on `glutin`/`glutin-winit` (the latter only
uses `x11`, and only forwards `wayland` to `glutin`) we may have dropped
a lot of "unused" dependencies for other GL backends, but also made the
crate unable to import X11 (Xlib/Xcb) and Wayland handles into EGL.
Also import the missing `glutin::context::Version` struct again which
was added last-minute to #6150 (to make sure my Intel card on Windows
creates a GLES 3.0+ instead of GLES 2.0 context) while the import was
accidentally squashed into #6152 (not merged yet).
* [wgpu-hal] Upgrade `glutin` to `0.31`
`glutin 0.30` onwards completely refactored its internals to no longer
be reliant on `winit`, as they (by default) have no direct relation
except needing to perform _some_ operations (platform-specific) at
strategic times in window creation and event loop handling. Most of
that is handled by a new `glutin-winit` introp crate, while the core
`glutin` crate now exclusively focuses on wrapping the various OpenGL
context APIs (CGL, EGL, WGL, ...).
This does result in a little more verbose handling to get the right
`GLDisplay`, `GLConfig`, `GLContext` and `GLSurface`, but gives much
more control and makes all intricacies more explicit. Most of the
code was copied from `glutin 0.31`'s example crate, with the code for
transparency support removed.
Note that the example doesn't at all handle event loop events properly:
resizes and redraws are not listened to, and mobile-specific surface
events (`Resumed` and `Suspended`) are equally ignored.
* [wgpu-hal] Implement proper `Surface` availability semantics in `raw-gles` example
This option was only evaluated for Metal backends, and now it's required
there so the option is going away. It is still configurable for tests
via the PipelineOptions struct, deserialized from .ron files.
This also fixes some type problems with the unpack functions in
writer.rs. Metal << operator extends operand to int-sized, which then
has to be cast back down to the real size before as_type bit conversion.
The math for the snorm values is corrected, in some cases using the
metal unpack_snorm2x16_to_float function because we can't directly
cast a bit-shifted ushort value to half.
It's already documented that to unmap a buffer it has to have been mapped.
Vulkan was the only backend that was returning an OOM on missing `Buffer.block` but `Buffer.map_buffer` already returns an error in this case.
* Expose gpu allocation configuration options
This commit adds hints to control memory allocations strategies to the configuration options. These hints allow for automatic profiles such as optimizing for performance (the default, makes sense for a game), optimizing for memory usage (typically more useful for a web browser or UI library) and specifying settings manually.
The details of gpu allocation are still in flux. The goal is to switch vulkan and metal to gpu_allocator which is currently used with d3d12. gpu_allocator will also likely receive more configuration options, in particular the ability to start with smaller memory block sizes and progressively grow the block size. So the manual settings already provision for this upcoming option. Another approach could be to wait and add the manual option after the dust settles.
The reason for providing presets and defining values in the backends is that I am convinced that optimal fonigurations should take hardware capabilities into consideration. It's a deep rabbithole, though, so that will be an exercise for later.
* changelog
* Update CHANGELOG.md
Co-authored-by: Andreas Reich <r_andreas2@web.de>
* Add a comment about not entirely knowing what we are doing
---------
Co-authored-by: Andreas Reich <r_andreas2@web.de>
Fix two major synchronization issues in `wgpu_val::vulkan`:
- Properly order queue command buffer submissions. Due to Mesa bugs, two semaphores are required even though the Vulkan spec says that only one should be necessary.
- Properly manage surface texture acquisition and presentation:
- Acquiring a surface texture can return while the presentation engine is still displaying the texture. Applications must wait for a semaphore to be signaled before using the acquired texture.
- Presenting a surface texture requires a semaphore to ensure that drawing is complete before presentation occurs.
Co-authored-by: Jim Blandy <jimb@red-bean.com>