Commit Graph
1188 Commits
Author SHA1 Message Date
teoxoy e10e17907f remove DeviceError::ResourceCreationFailed 2025-07-02 11:11:37 +02:00
teoxoy 918fdc98b5 remove reason from ErrorType::DeviceLost since it's always "unknown" and users should register the lost callback instead 2025-07-02 11:11:37 +02:00
teoxoy 25d75e2a02 remove DeviceError::Invalid, this variant should be the same as DeviceError::Lost 2025-07-02 11:11:37 +02:00
Andy Leiserson 86e6b1835b Tolerate destruction of textures used in immediate queue operations prior to submit 2025-06-30 10:29:50 -07:00
Erich Gubler 964c5bbf4c feat(core): impl WebGpuError for API errors 2025-06-30 10:58:56 -04:00
Andy Leiserson 09c08037c9 Add a field in BasePass to store errors 2025-06-26 15:39:53 -07:00
SupaMaggie70IncorporatedandConnor Fitzgerald e40e66d205 DXIL & HLSL passthrough (#7831)
Co-authored-by: Connor Fitzgerald <connorwadefitzgerald@gmail.com>
2025-06-25 18:12:58 -04:00
Vecvec dc0d457a9a Add limits for acceleration structures (#7845) 2025-06-25 14:49:55 -04:00
Jamie Nicol 8cdbcc1755 Add ExternalTexture BindingType behind new Feature flag
Adds a new feature flag, `EXTERNAL_TEXTURE`, indicating device support
for our implementation of WebGPU's `GPUExternalTexture` [1] which will
land in upcoming patches. Conceptually this would make more sense as a
downlevel flag, as it is a core part of the WebGPU spec which we do not
yet support. We do not want, however, to cause applications to reject
adapters because we have not finished implementing this, so for now we
are making it an opt-in feature.

As an initial step towards supporting this feature, this patch adds a
new `BindingType` corresponding to WebGPU's
`GPUExternalTextureBindingLayout` [2]. This binding type dictates that
when creating a bind group the corresponding entry must be either an
external texture or a texture view with certain additional requirements
[3].

As of yet wgpu has no concept of an external texture (that will follow
in later patches) but for now this patch ensures that texture views
corresponding to an external texture binding type are validated
correctly. Note that as the feature flag is not yet supported on any
real backends, bind group layout creation will fail before getting the
chance to attempt to create a bind group. But in the added tests using
the noop backend we can see this validation taking place.

[1] https://www.w3.org/TR/webgpu/#gpuexternaltexture
[1] https://www.w3.org/TR/webgpu/#dictdef-gpuexternaltexturebindinglayout
[2] https://gpuweb.github.io/gpuweb/#bind-group-creation
2025-06-17 15:00:29 -07:00
Dmitry Zamkov bbb7cc79ef Implement clip-distances extension for GL and Vulkan backends (#7730)
* Basic implementation of `clip_distances` for Vulkan and GL backends

* Added GPU test for `clip-distances`

* Update feature array size

* Add changelog entry

* Validate `clip_distances` array size

* Check for `clip_distances` enable directive

* Consolidate code for generating `enable` directives in WGSL backend and add `clip_distances`.
2025-06-16 10:33:31 +02:00
Vecvec aa427de4ea Remove unsafe acceleration structure build (#7513) 2025-06-12 00:43:03 -04:00
VecvecandConnor Fitzgerald 73eb83ded1 [wgpu] Blas compaction (#7285)
Co-authored-by: Connor Fitzgerald <connorwadefitzgerald@gmail.com>
2025-06-11 21:40:10 -04:00
Andy Leiserson ce89c916f8 Add storage for an error in CommandEncoderStatus (#7785) 2025-06-11 09:56:14 -07:00
Andy Leiserson 3cca5f8cfd Make the copy_buffer_to_buffer size parameter optional (#7659)
* Make wgpu-core's copy_buffer_to_buffer `size` parameter optional

* Make the copy size optional in more places

* Fix for webgpu backend

* [deno_webgpu] Support additional copyBufferToBuffer signatures

* Add changelog entry

* Add copyBufferToBuffer tests to CTS test list

(This doesn't actually enable the tests for the new overloads, because
of a different error reporting issue that affects many CTS tests
including these. But if you run the tests for the new overloads
manually, before and after the fix, you can see that the behavior has
changed.)

* Reproducible formula for vendoring modified webgpu-sys

Commit the updated vendor command in all the files for consistency.
2025-05-28 17:21:19 +02:00
Andy Leiserson 24d0eae36c Remove another type of error that could be raised by destroy (#7720) 2025-05-24 17:02:00 +02:00
Mehmet Oguz Derin 2a62299a84 Support Sliced 3D for ASTC (#7577)
Enables "texture-compression-astc-sliced-3d" for backends that support ASTC (such as excluding D3D12).
2025-05-04 12:48:16 +02:00
teoxoy d14849df48 [vk] add support for rendering to slices of 3D textures 2025-04-25 12:39:45 +02:00
teoxoy d714e3d95a [d3d12,metal,gl] add support for rendering to slices of 3D textures 2025-04-25 12:39:45 +02:00
teoxoy 6058dd4296 [wgpu-core] allow 2D-Array texture views with a single layer to be used as render attachments 2025-04-25 12:39:45 +02:00
Teodor Tanasoaia 195217b269 allow read-write and read-only storage textures without TEXTURE_ADAPTER_SPECIFIC_FORMAT_FEATURES (#7623) 2025-04-24 11:45:24 -04:00
teoxoy 7b482622d3 allow sampler creation to return OOMs
This is to preserve the current behavior as tested by the `SAMPLER_CREATION_FAILURE` test.

This is not spec compliant but it's unclear what we should do instead. I opened https://github.com/gpuweb/gpuweb/issues/5142 to figure out what we should do.
2025-04-22 09:21:27 +02:00
teoxoy c13fc3d141 don't call release_gpu_resources when losing the device
This removes the possibility of deadlocks happening since `release_gpu_resources` tries to lock resources (trackers, snatchable_lock, pending_writes, life_tracker) while they might be already locked; `handle_hal_error` is called in lots of places.

Removing the call only delays destruction since `release_gpu_resources` is still called in `maintain`.
2025-04-22 09:21:27 +02:00
teoxoy ece29b6e68 [D3D12/VK] add OOM check on submit and poll that will lose the device if we are over 95% of our budget 2025-04-22 09:21:27 +02:00
teoxoy c144f2a697 invalidate Device on OOM errors with the exception of buffer, texture, query set and acceleration structure creation 2025-04-22 09:21:27 +02:00
teoxoy c49620d1a2 add device validity checks to Queue methods 2025-04-22 09:21:27 +02:00
RedMindZ f64eae7734 Fixed a deadlock caused by locking the device's snatchable lock **after** locking the queue's pending writes (#7582) 2025-04-21 11:08:57 -04:00
teoxoy 597114052e [d3d12] get vertex_index & instance_index builtins working for indirect draws 2025-04-15 17:56:17 +02:00
Vecvec 8010203281 Rework acceleration structure build tracking. 2025-04-09 22:27:04 -04:00
Vecvec 382a1e3c9b Make submission index lockable. 2025-04-09 22:27:04 -04:00
Sylvain BennerandConnor Fitzgerald 14690470bb [metal] Metal compute shader passthrough (#7326)
Co-authored-by: Connor Fitzgerald <connorwadefitzgerald@gmail.com>
2025-04-09 14:25:41 +00:00
Connor Fitzgerald c860a2cf8c Rename and unsafe start_capture -> start_graphics_debugger_capture (#7470)
* Improve `start_capture` docs

* Docs
2025-04-03 13:07:22 -04:00
Connor Fitzgerald c344bec508 Comments 2025-04-03 10:35:20 -04:00
Connor Fitzgerald 6a986f4bc4 Timestamp normalization 2025-04-03 10:35:20 -04:00
Connor Fitzgerald a9279de793 Fix assorted issues with WebGL (#7448)
* Fix validation error when configuring the surface on WebGL

* Remove unneeded `webgl` feature

* Fix compilation of the `noop` backend on `wasm32`

* Prevent `webgpu` examples from incorrectly falling back to WebGL

* Reduce dependency set when building wasm examples

* Fix various warnings
2025-03-31 22:02:21 -04:00
teoxoy 7d15c511de move indirect validation code to its own folder
rename items appropriately
internalize indirect buffer bind groups
2025-03-26 18:26:31 +00:00
teoxoy 0285e60984 Replace the indirect-validation feature with InstanceFlags::VALIDATION_INDIRECT_CALL.
With the only caveat that device creation will now panic if the `wgsl` feature is not enabled, `InstanceFlags::VALIDATION_INDIRECT_CALL` is set and the device supports `DownlevelFlags::INDIRECT_EXECUTION`.
2025-03-26 18:26:31 +00:00
teoxoy 53c2d15e0b implement indirect draw validation 2025-03-26 18:26:31 +00:00
Bruce Mitchener ccaf9a84f8 Fix clippy::implicit_saturating_sub lints (#7395) 2025-03-22 13:17:34 -04:00
Connor Fitzgerald 4687973c9a Use bytemuck instead of slice::from_raw_parts for transmutes (#7376) 2025-03-19 22:05:21 -04:00
c6286791fe feat: implement F16 support in shaders (#5701)
Co-authored-by: FL33TW00D <fleetwoodpersonal@gmail.com>
Co-authored-by: Connor Fitzgerald <connorwadefitzgerald@gmail.com>
Co-authored-by: ErichDonGubler <erichdongubler@gmail.com>
2025-03-19 16:26:50 +00:00
David ZhongandErich Gubler 3eb1bd53bb Use TextureViewDescriptor::usage when Creating Texture View (#7162)
Co-authored-by: Erich Gubler <erichdongubler@gmail.com>
2025-03-11 13:01:23 -04:00
Kevin Reid 326ad03ce1 Move trace_dir/trace_path to a custom enum inside DeviceDescriptor.
This allows `wgpu` to not unconditionally depend on `std::path::Path`.
It’s also, in my opinion, more user-friendly, because the feature which
most users will not use (and is not currently functional) is now a
defaultable struct field instead of a required parameter.

The disadvantage is that `wgpu-types` now has to know about tracing.
2025-03-10 22:17:06 -04:00
Kevin Reid 16d41cfaf3 Make std usage optional for wgpu-core. (#7279)
* Make `wgpu_core::snatch::LockTrace` fully instead of partially conditional.

Now, when `cfg(not(debug_assertions))`, there is no `SNATCH_LOCK_TRACE`
thread local, and `LockTrace` has no data fields.

* Make `std` usage optional for `wgpu-core`.

Adds a `std` feature, enabled by default, to `wgpu-core`. When that
feature is disabled, the following functionality is not available:

* `Send + Sync` for resources.
* `trace` feature.
* `observe_locks` feature.
* Snatch lock recursive locking assertion.
2025-03-06 14:07:38 -05:00
Vecvec 5b3266db23 Support getting hit vertex positions (#7183) 2025-03-04 20:06:44 +01:00
Erich Gubler 6f0c2434a7 chore(hal,core): use core::{error::Error,format} instead of std
This allows `wgpu-hal` to be used in `no_std` environments, except that
currently, only the `noop` backend supports `no_std`.
In the future, `cfg(all(gles, webgl))` should also be `no_std`, but that
requires further work.

Co-Authored_by: Kevin Reid <kpreid@switchb.org>
2025-02-26 22:58:50 -05:00
Flávio J. SaraivaandConnor Fitzgerald a26171b2a0 Fix compilation for targets that don't have AtomicU64. (#7118)
Co-authored-by: Connor Fitzgerald <connorwadefitzgerald@gmail.com>
2025-02-24 22:07:13 +00:00
Kevin Reid 7bcbfe0712 wgpu-core: Adjust imports to allow no_std.
Dependencies on `std` that still exist:
* Locks
* `std::error::Error` (waiting for Rust 1.81)
* `std::os::raw::c_char` for FFI
* `thread_local` and `Backtrace` in `snatch`
2025-02-21 08:47:16 +01:00
Connor Fitzgerald 194d4b1f36 New Binding Array Limit (#6952) 2025-02-17 09:22:24 -05:00
Connor Fitzgerald d8833d0798 Use Update After Bind Descriptors for Bind Groups With Binding Arrays (#6815)
* Use Update After Bind Descriptors for Bind Groups With Binding Arrays

Update After Bind

x

* Comments

* Fix URL
2025-02-15 12:02:27 -05:00
Connor Fitzgerald 7e119968ce Properly Deal with Timeouts (#7030) 2025-02-14 23:19:51 +00:00