webgpu: Update wgpu and revamp computepass (#32575)

* Do not wait on drop, but rather wake poller thread

* Update wgpu and render stuff

* Set some good expectations

* Update wgpu again

* handle IPC error as warning

* More good expectations

* Some more expectations

CTS does not match the spec: https://github.com/gpuweb/cts/issues/3806

* This expectations are due to other changes in servo

also happening on main

* Explain error_command_encoders and remove RefCell around it

* fixup

* store validness of passes

* More good expectations

* More docs

* this assert is wrong

* This is even more right per CTS/spec

Only Command encoder state errors are allowed here, but wgpu does not exposes them.

* More good expectations

* One bad expectation

* Fix my english
This commit is contained in:
Samson 2024-06-28 06:49:35 +02:00 committed by GitHub
parent fced0b4940
commit e9cf4d4971
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
16 changed files with 372 additions and 1870 deletions

View file

@ -19,7 +19,7 @@ use crate::dom::bindings::str::USVString;
use crate::dom::globalscope::GlobalScope;
use crate::dom::gpubindgroup::GPUBindGroup;
use crate::dom::gpubuffer::GPUBuffer;
use crate::dom::gpucommandencoder::{GPUCommandEncoder, GPUCommandEncoderState};
use crate::dom::gpucommandencoder::GPUCommandEncoder;
use crate::dom::gpurenderbundle::GPURenderBundle;
use crate::dom::gpurenderpipeline::GPURenderPipeline;
@ -164,16 +164,12 @@ impl GPURenderPassEncoderMethods for GPURenderPassEncoder {
let render_pass = self.render_pass.borrow_mut().take();
self.channel
.0
.send(WebGPURequest::RunRenderPass {
command_encoder_id: self.command_encoder.id().0,
.send(WebGPURequest::EndRenderPass {
render_pass,
device_id: self.command_encoder.device_id().0,
})
.expect("Failed to send RunRenderPass");
self.command_encoder.set_state(
GPUCommandEncoderState::Open,
GPUCommandEncoderState::EncodingRenderPass,
);
Ok(())
}