Commit Graph
999 Commits
Author SHA1 Message Date
teoxoy 781b54a8b9 remove TrackingData from resources that are not tracked 2024-08-08 14:42:13 +02:00
teoxoy 8c7c5c4974 decouple device and queue IDs
Devices and queues can have different lifetimes, we shouldn't assume that their IDs match.
2024-08-05 15:46:12 +02:00
teoxoy 4e777bd0e7 merge the texture and texture view trackers of BindGroupStates 2024-08-03 11:58:14 +02:00
teoxoy 62af9d78b5 rename {Buffer,Texture}BindGroupState's add_single to insert_single
Also change it's definition to take an owned `Arc`. This makes these functions consistent with the other trackers.
2024-08-03 11:58:14 +02:00
teoxoy 5cb1be63aa refactor the StatelessTracker to hold a Vec of Arcs
Also removes the `StatelessBindGroupState` since it does the same thing.
2024-08-03 11:58:14 +02:00
teoxoy 3a5ad193db remove all internal BindGroupState Mutexes 2024-08-03 11:58:14 +02:00
Nicolas Silva 9c6ae1beae Log spring cleaning (#6065)
Remove unused logs in wgpu-core and wgpu-hal
2024-07-31 16:24:23 +00:00
AthosOfAthos bfad205cf5 Reduced verbosity for 'Device::maintain: waiting for submission index… (#6044) 2024-07-30 10:58:38 +02:00
teoxoy 650054bbcd [player] simplify sync buffer writing 2024-07-29 17:31:24 +02:00
teoxoy 7502e65213 remove unused Global.device_get_buffer_sub_data 2024-07-29 17:31:24 +02:00
teoxoy 2ea081fabf remove waiting functionality from Global.{buffer,texture,texture_view}_drop()
Those resources won't be destroyed if used by a submission anyway.
2024-07-29 17:31:24 +02:00
Samson 7462754bde Remove 'de: 'static" serde bound and replace &'static str with Cow in some errors (#6048)
* Remove `serde(bound(deserialize = "'de: 'static"))` and replace `&'static str` with `Cow` in deser errors

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

* Allow `clippy::result_large_err`

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

---------

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>
2024-07-29 09:07:26 +02:00
Andreas Reich d3c38a4fd0 Fix AnyDevice drop implementation dropping the wrong thing (#6052) 2024-07-26 10:31:36 +02:00
teoxoy 03f6d24ea6 deduplicate derived BGLs 2024-07-25 18:15:29 +02:00
Erich Gubler 65b6e15f0f chore: satisfy clippy::unused_qualifications 2024-07-24 12:09:15 -04:00
Erich Gubler 723995d9a9 refactor: warn on and satisfy clippy::{ptr_as_ptr,ref_as_ptr} in wgpu-{core,hal,types}
…using `cargo +1.79.0 clippy --workspace --all-features --all-targets
--fix`, plus some manual changes to (1) catch some missed cases (not run
on all platforms?) and (2) `--fix` doesn't make things compile again. 😀
2024-07-24 12:09:15 -04:00
Jim Blandy ebb011fc6b [core] Use ManuallyDrop for wgpu_core::device::Queue::raw.
Change the field `wgpu_core::device::Queue::raw` from an
`Option<A::Queue>` to a `std::mem::ManuallyDrop<A::Queue>`. Replace
various `.as_ref().unwrap()` chains with calls to a new accessor
function `Queue::raw`.

An `Option` is misleading, as this field is always populated during
the lifetime of a `Queue`. Instead, we simply have a field whose value
needs to be moved in `<Queue as Drop>::drop`; `ManuallyDrop` is the
Rust idiom for this situation.
2024-07-23 14:02:16 -07:00
teoxoy 205f1e3ab6 [wgpu-core] fix length of copy in queue_write_texture #2
Follow-up to 6f16ea460a & 7e112ca4c0.
2024-07-22 12:25:07 +02:00
Vladas Zakrevskis 5a0e2187f0 Print requested and supported usages on UnsupportedUsage error (#6007)
* Print requested and supported usages on UnsupportedUsage error

* fmt

* changelog
2024-07-21 21:12:28 -04:00
sagudev fc87033b30 Make BufferAccessError (de)serializable
Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>
2024-07-19 13:37:57 -04:00
Brad Werth 6cd387412f Remove vertex_pulling_transfrom from PipelineCompilationOptions.
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.
2024-07-19 17:13:45 +02:00
Nicolas Silva bc7622f641 Expose GPU allocation reports in wgpu, wgpu-core and wgpu-hal 2024-07-19 10:16:29 +02:00
Connor Fitzgerald 3c3b532cf3 Remove latest_submission_index (#5976)
* Remove latest_submission_index

* CI

* Comments
2024-07-18 16:43:39 +00:00
teoxoy 9a0adefe88 use ManuallyDrop instead of Option for PendingWrites 2024-07-18 10:30:11 -04:00
teoxoy 77e45d46df add missing device valid check in create_texture_view 2024-07-18 10:29:08 -04:00
teoxoy f767220399 remove same device check from create_texture_binding
`view.parent` will have the same `device` as the `view` itself
2024-07-18 10:29:08 -04:00
teoxoy 63303d4b4e add missing same device checks 2024-07-18 10:29:08 -04:00
Jim Blandy aeb2067e81 [core] Make poll(Wait) not hang after bad command submission.
Add `wgpu_core::device::Device::last_successful_submission_index`,
which records the fence value that `Maintain::Wait` should actually
wait for. See comments for details.

Fixes #5969.
2024-07-17 16:11:04 -07:00
Jim Blandy 2bc328c46f [hal, core] Introduce wgpu_hal::AtomicFenceValue, and use it.
Introduce the new type alias `wgpu_hal::AtomicFenceValue`, which is
the atomic version of `wgpu_hal::FenceValue`. Use this type alias in
`wgpu_core`. Remove `as` conversions made unnecessary since we're not
conflating `usize` with `u64` any more.
2024-07-17 16:11:04 -07:00
teoxoy 91924fb603 [wgpu-core] make implicit_pipeline_ids arg optional for users that don't provide IDs 2024-07-17 16:57:01 +02:00
Teodor Tanasoaia 7e112ca4c0 [wgpu-core] fix length of copy in queue_write_texture (#5973)
The size of the given `data` might be less than the size of the staging buffer.
This issue became apparent with the refactor in 6f16ea460a (https://github.com/gfx-rs/wgpu/pull/5946) since there is now an assert in `StagingBuffer.write()`.

Ruffle ran into this in https://github.com/gfx-rs/wgpu/issues/3193#issuecomment-2231209711.
2024-07-17 10:33:48 -04:00
Jim Blandy d02e2949b2 [core] Correctly check mipmap-filtering samplers against the layout.
Ensure that samplers using non-`Nearest` mipmap filtering are
considered "filtering samplers" when deciding bind group layout
compatibility.

Add tests for layout `NonFiltering` validation.

Fixes #5948.
2024-07-15 10:03:49 +02:00
Jim Blandy 05c0656fa4 [core] Correct docs for LifetimeTracker and its mapped field. 2024-07-15 09:59:04 +02:00
teoxoy 6f16ea460a make the StagingBuffer implementation more robust 2024-07-12 16:33:57 +02:00
teoxoy 347d902bcb introduce FlushedStagingBuffer 2024-07-12 16:33:57 +02:00
teoxoy 2f282cdd06 remove instance_flags arg from StagingBuffer::new 2024-07-12 16:33:57 +02:00
teoxoy 26f65ddffd [wgpu-hal] remove return type from Device.unmap_buffer()
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.
2024-07-12 16:33:57 +02:00
teoxoy 5e2df1406d use StagingBuffer.flush() in Buffer.unmap_inner()
We should have always unmapped the staging buffer as we are using it on the GPU after this point.
2024-07-12 16:33:57 +02:00
teoxoy 9a7f44bf09 make StagingBuffer creation a method 2024-07-12 16:33:57 +02:00
teoxoy a8b0f2f6a6 remove device arg from StagingBuffer.flush() 2024-07-12 16:33:57 +02:00
teoxoy 4d285d8b61 remove the Mutex around StagingBuffer's internal buffer 2024-07-12 16:33:57 +02:00
teoxoy a0c185a28c [wgpu-core] fix trying to create 0 sized staging buffers when creating mapped_at_creation buffers
This issue was introduced by fabbca294a.
2024-07-11 14:10:22 +02:00
teoxoy ef0ce05d3a [wgpu-core] fix .zip() usages 2024-07-10 16:42:35 +02:00
teoxoy 425526828f introduce DeviceTextureTracker which holds weak references to textures 2024-07-09 10:48:52 +02:00
teoxoy aa9cb71a54 introduce DeviceBufferTracker which holds weak references to buffers 2024-07-09 10:48:52 +02:00
teoxoy 0e1c1f7c07 replace the tracker in Device with a new DeviceTracker 2024-07-09 10:48:52 +02:00
teoxoy 3142e15907 remove the triage suspected machinery 2024-07-09 10:48:52 +02:00
teoxoy 3fba403060 don't check if the buffer is still "present" from the user's perspective in LifetimeTracker.handle_mapping
This change doesn't change behavior as `Global.buffer_drop` already unmaps the buffer.
2024-07-09 10:48:52 +02:00
teoxoy 3cc6c2743a remove LifetimeTracker.future_suspected_{buffers,textures} 2024-07-09 10:48:52 +02:00
teoxoy 61739d9583 remove PendingWrites.executing_command_buffers
The Vec only ever contained 0 or 1 command buffers.
We now acquire an encoder on every submit for pending writes but that's not a problem since those are coming from a pool anyway.
2024-07-09 10:48:52 +02:00