Commit Graph
1068 Commits
Author SHA1 Message Date
teoxoy 94e040bc8a avoid creating the bind group for indirect validation if buffer size is 0 2024-10-17 18:41:24 +02:00
teoxoy bbee35b145 [d3d12] get num_workgroups builtin working for indirect dispatches 2024-10-14 15:22:18 +02:00
Teodor Tanasoaia 7f708edd1f Ensure safety of indirect dispatch (#5714)
by injecting a compute shader that validates the content of the indirect buffer
2024-10-14 15:02:01 +02:00
teoxoy 952ba3e12c move queue methods on the Queue type 2024-10-14 13:24:44 +02:00
teoxoy 321bd64ba2 add Adapter::backend() 2024-10-14 13:24:44 +02:00
Erich Gubler 91447aefc9 refactor: s/once_cell::sync::OnceCell/std::sync::OnceLock
Weaken our dependence on the `once_cell` crate by using functionality
from `std` instead that was upstreamed from `once_cell`.

It's not yet possible to eliminate this dependency entirely, but do what
we can for now.
2024-10-11 15:52:21 -07:00
teoxoy 38a13b94f0 refactor: make Snatchable::snatch take _guard by &mut _ 2024-10-02 04:41:34 -04:00
Erich Gubler 0903ba6d47 refactor(core): delete unused CreateDeviceError 2024-10-02 04:41:34 -04:00
Elie Michel c9202ee54a Remove redundant let binding (#6356) 2024-10-02 10:21:30 +02:00
Erich Gubler fc2fd95a98 fix: handle Queue::submit non-fatally (#6318)
* Change the signature of `wgpu_core::Global::queue_submit` to return
  a `(SubmissionIndex, …)` in addition to its current error type.
* Change the control flow of errors in `Queue::submit` to break to the
  end of a block. This is similar to what we already do in many APIs in
  `wgpu_core`.
* Hoist the scope of the local `submit_index` binding so it can be used
  at the point where we need to convert current error paths to also
  return the submission index.

Later, we will likely want to avoid actually retrieving a new submission
index so we can minimize the critical section of code. We'll need to
figure out a strategy for returning a valid (but not necessarily unique)
index in the case of failures that prevent successful submission.
2024-09-25 02:52:25 +00:00
Erich Gubler 841e7c8857 fix: only log Device::maintain waits if they happen (#6303) 2024-09-21 18:47:35 +02:00
Teodor Tanasoaia c8beade187 remove backend from ID (#6263) 2024-09-13 17:30:38 +02:00
teoxoy 2addb267fc [wgpu-core] remove FutureId.into_id() 2024-09-10 11:58:54 +02:00
teoxoy c182d3f24f [wgpu-core] rename .strict_unregister() & .strict_remove() to .remove() 2024-09-10 11:58:54 +02:00
teoxoy ac74f36590 [wgpu-core] rename .strict_get() to .get() 2024-09-10 11:58:54 +02:00
teoxoy 31edbfd0dc [wgpu-core] use .strict_get() & .strict_unregister() for command buffers
We now only rely on the presence of the inner `CommandBufferMutable` to determine if command buffers are invalid.
2024-09-10 11:58:54 +02:00
teoxoy c0c594eff2 [wgpu-core] use Fallible for RenderBundle 2024-09-10 11:58:54 +02:00
teoxoy ee3b63a9af [wgpu-core] use Fallible for PipelineCache 2024-09-10 11:58:54 +02:00
teoxoy 5bcdd4266e [wgpu-core] use Fallible for ShaderModule 2024-09-10 11:58:54 +02:00
teoxoy 1c5793afb3 [wgpu-core] use Fallible for PipelineLayout 2024-09-10 11:58:54 +02:00
teoxoy e444e78d94 [wgpu-core] use Fallible for BindGroupLayout 2024-09-10 11:58:54 +02:00
teoxoy 45206e553d [wgpu-core] use Fallible for RenderPipeline 2024-09-10 11:58:54 +02:00
teoxoy 82ce2ea747 [wgpu-core] use Fallible for ComputePipeline 2024-09-10 11:58:54 +02:00
teoxoy b42500ee40 [wgpu-core] use Fallible for BindGroup 2024-09-10 11:58:54 +02:00
teoxoy 2ea314e978 [wgpu-core] remove Arc around StagingBuffer 2024-09-10 11:58:54 +02:00
teoxoy b14fe14e88 [wgpu-core] use Fallible for QuerySet 2024-09-10 11:58:54 +02:00
teoxoy 3ed1abc492 [wgpu-core] use Fallible for Sampler 2024-09-10 11:58:54 +02:00
teoxoy 04f17d4197 [wgpu-core] use Fallible for TextureView 2024-09-10 11:58:54 +02:00
teoxoy c630821f1d [wgpu-core] use Fallible for Texture 2024-09-10 11:58:54 +02:00
teoxoy 68e8b7d4a8 [wgpu-core] introduce Fallible and use it for Buffer (first step of invalidity internalization) 2024-09-10 11:58:54 +02:00
teoxoy 3437589107 [wgpu-core] use .strict_get() & .strict_unregister() for devices
This works because we never assign errors to devices (they are never invalid).
2024-09-10 11:58:54 +02:00
teoxoy 07c397a1e5 [wgpu-core] use .strict_get() & .strict_unregister() for queues
This works because we never assign errors to queues (they are never invalid).
2024-09-10 11:58:54 +02:00
teoxoy da9afea271 [wgpu-core] use .strict_get() & .strict_unregister() for surfaces
This works because we never assign errors to surfaces (they are never invalid).
2024-09-10 11:58:54 +02:00
teoxoy 98426329a4 [wgpu-core] introduce Registry .strict_get() & .strict_unregister() and use them for adapters
This works because we never assign errors to adapters (they are never invalid).
2024-09-10 11:58:54 +02:00
teoxoy 70a9c01b48 [wgpu-core] make the Registry generic over T: Clone 2024-09-10 11:58:54 +02:00
teoxoy 480abae331 fix ordering of device.valid.store in device_destroy 2024-09-09 15:03:19 +02:00
teoxoy eb47449eb9 invalidate the device when we encounter driver-induced device loss or on unexpected errors 2024-09-09 15:03:19 +02:00
Nicolas Silva 81963e24ca Fix the internal buffer and texture count (#6223)
Currently, we only increment the internal buffer/texture counters when creating them in the regular way (not when creating them from externally built hal objects (create_texture_from_hal/create_buffer_from_hal). However we decrement the counter in all cases, which makes the counters incorrect when these externally created resources are involved.

This commit fixes it by adding hooks (add_raw_buffer and add_raw_texture) in the hal device abstractions to inform when buffer or textures are created externally.
2024-09-05 18:53:01 +00:00
Teodor Tanasoaia 4e78829d82 [wgpu-core] when mapping buffers for reading, mark buffers as initialized only when they have MAP_WRITE usage (#6178) 2024-09-04 14:57:08 +00:00
Jim Blandy ee35b0e586 [core, hal, types] Clarify wgpu_hal's bounds check promises.
In `wgpu_hal`:

- Document that `wgpu_hal` guarantees that shaders will not access buffer
  contents beyond the bindgroups' bound regions, rounded up to some
  adapter-specific alignment. Introduce the term "accessible region" for
  the portion of the buffer that shaders can actually get at.

- Document that all bets are off if you disable bounds checks with
  `ShaderModuleDescriptor::runtime_checks`.

- Provide this alignment in `wgpu_hal::Alignments`. Update all backends
  appropriately.

- In the Vulkan backend, use Naga to inject bounds checks on buffer accesses
  unless `robustBufferAccess2` is available; `robustBufferAccess` is not
  sufficient. Retrieve `VK_EXT_robustness2`'s properties, as needed to discover
  the alignment above.

In `wgpu_core`:

- Use buffer bindings' accessible regions to determine which parts of the buffer
  need to be initialized.

In `wgpu_types`:

- Document some of the possible effects of using
  `ShaderBoundsChecks::unchecked`.

Fixes #1813.
2024-09-03 12:33:02 -04:00
Jim Blandy de7765bd28 [core]: Let Device::create_buffer_binding get limits from self.
Rather than passing `self.limits` to `Device::create_buffer_binding`
as an argument, let it simply refer to `self.limits` itself.
2024-09-03 12:33:02 -04:00
Jim Blandy 9ebc50fc1e [core] Remove unnecessary pub(crate) from Device methods.
Remove the `pub(crate)` visibility marking from various associated
functions of `Device` that are defined in, and not used outside of,
the `wgpu_core::device::resource` module.
2024-09-03 12:33:02 -04:00
Jim Blandy 9d24406d29 [core] Simplify self types in device::resource.
Change various functions that have no need to create an owning
reference to the `Device` to accept `&self` instead of `&Arc<Self>`.

Change `ParentDevice::same_device` to accept `&Device` as the point of
comparison, not `&Arc<Device>`. Call sites will use Deref conversion,
so no callers need to be changed.
2024-09-03 12:33:02 -04:00
Jim Blandy 504ff3e1a6 [core] Rename map_buffer to resolve_buffer, etc.
In `Device::create_bind_group`, name the functions that convert
resource ids to Arcs `resolve_foo`, not `map_foo`:

- The types that hold Arcs are usually called `ResolvedBlah`.

- The name `map_buffer` is misleading.
2024-09-03 12:33:02 -04:00
Jim Blandy bbdbafdf8a [core] Add lock::observing module, for analyzing lock acquisition.
Add a new module `lock::observing`, enabled by the `observe-locks`
feature, that records all nested lock acquisitions in trace files.

Add a new utility to the workspace, `lock-analyzer`, that reads the
files written by the `observe-locks` feature and writes out a new
`define_lock_ranks!` macro invocation that covers all observed lock
usage, along with comments giving the held and acquired source
locations.
2024-08-28 22:44:20 -04:00
teoxoy 2c31414517 add an Unexpected variant to DeviceError 2024-08-15 13:56:05 +01:00
Andreas Reich 9794f33823 Handle webgl's queue_copy_external_image_to_texture 2024-08-14 17:34:11 +02:00
Andreas Reich 24498f04d4 The second unraveling: hub and all types on it are generic free!
gfx_select macros are empty husks now that are waiting to be removed
2024-08-14 17:34:11 +02:00
Andreas Reich 3181251577 Core adapter no longer uses any generics 2024-08-14 17:34:11 +02:00
Andreas Reich 04cadfb369 Core's Surface, Instance and Adapter use now dynamic hal types 2024-08-14 17:34:11 +02:00