mirror of
https://github.com/servo/servo.git
synced 2025-06-06 16:45:39 +00:00
* Prepare errorscopes logic in wgpu_thread * remove scope_id from ipc * new GPUErrors per spec * remove cotent timeline error_scope * fixup poperrorscope types * device_scope -> gpu_error and nice errors * Handle errors detection more elegantly * good expectations * new expectations * Make error_scope.errors Vec as per spec
224 lines
7.2 KiB
Rust
224 lines
7.2 KiB
Rust
/* This Source Code Form is subject to the terms of the Mozilla Public
|
|
* License, v. 2.0. If a copy of the MPL was not distributed with this
|
|
* file, You can obtain one at https://mozilla.org/MPL/2.0/. */
|
|
|
|
use std::rc::Rc;
|
|
|
|
use dom_struct::dom_struct;
|
|
use ipc_channel::ipc::IpcSharedMemory;
|
|
use webgpu::{wgt, WebGPU, WebGPUQueue, WebGPURequest, WebGPUResponse};
|
|
|
|
use super::bindings::codegen::Bindings::WebGPUBinding::{GPUImageCopyTexture, GPUImageDataLayout};
|
|
use super::gpu::{response_async, AsyncWGPUListener};
|
|
use crate::dom::bindings::cell::DomRefCell;
|
|
use crate::dom::bindings::codegen::Bindings::WebGPUBinding::{
|
|
GPUExtent3D, GPUQueueMethods, GPUSize64,
|
|
};
|
|
use crate::dom::bindings::codegen::UnionTypes::ArrayBufferViewOrArrayBuffer as BufferSource;
|
|
use crate::dom::bindings::error::{Error, Fallible};
|
|
use crate::dom::bindings::reflector::{reflect_dom_object, DomObject, Reflector};
|
|
use crate::dom::bindings::root::{Dom, DomRoot};
|
|
use crate::dom::bindings::str::USVString;
|
|
use crate::dom::globalscope::GlobalScope;
|
|
use crate::dom::gpubuffer::{GPUBuffer, GPUBufferState};
|
|
use crate::dom::gpucommandbuffer::GPUCommandBuffer;
|
|
use crate::dom::gpuconvert::{
|
|
convert_ic_texture, convert_image_data_layout, convert_texture_size_to_dict,
|
|
convert_texture_size_to_wgt,
|
|
};
|
|
use crate::dom::gpudevice::GPUDevice;
|
|
use crate::dom::promise::Promise;
|
|
|
|
#[dom_struct]
|
|
pub struct GPUQueue {
|
|
reflector_: Reflector,
|
|
#[ignore_malloc_size_of = "defined in webgpu"]
|
|
#[no_trace]
|
|
channel: WebGPU,
|
|
device: DomRefCell<Option<Dom<GPUDevice>>>,
|
|
label: DomRefCell<USVString>,
|
|
#[no_trace]
|
|
queue: WebGPUQueue,
|
|
}
|
|
|
|
impl GPUQueue {
|
|
fn new_inherited(channel: WebGPU, queue: WebGPUQueue) -> Self {
|
|
GPUQueue {
|
|
channel,
|
|
reflector_: Reflector::new(),
|
|
device: DomRefCell::new(None),
|
|
label: DomRefCell::new(USVString::default()),
|
|
queue,
|
|
}
|
|
}
|
|
|
|
pub fn new(global: &GlobalScope, channel: WebGPU, queue: WebGPUQueue) -> DomRoot<Self> {
|
|
reflect_dom_object(Box::new(GPUQueue::new_inherited(channel, queue)), global)
|
|
}
|
|
}
|
|
|
|
impl GPUQueue {
|
|
pub fn set_device(&self, device: &GPUDevice) {
|
|
*self.device.borrow_mut() = Some(Dom::from_ref(device));
|
|
}
|
|
}
|
|
|
|
impl GPUQueueMethods for GPUQueue {
|
|
/// <https://gpuweb.github.io/gpuweb/#dom-gpuobjectbase-label>
|
|
fn Label(&self) -> USVString {
|
|
self.label.borrow().clone()
|
|
}
|
|
|
|
/// <https://gpuweb.github.io/gpuweb/#dom-gpuobjectbase-label>
|
|
fn SetLabel(&self, value: USVString) {
|
|
*self.label.borrow_mut() = value;
|
|
}
|
|
|
|
/// <https://gpuweb.github.io/gpuweb/#dom-gpuqueue-submit>
|
|
fn Submit(&self, command_buffers: Vec<DomRoot<GPUCommandBuffer>>) {
|
|
let valid = command_buffers.iter().all(|cb| {
|
|
cb.buffers()
|
|
.iter()
|
|
.all(|b| matches!(b.state(), GPUBufferState::Unmapped))
|
|
});
|
|
if !valid {
|
|
self.device
|
|
.borrow()
|
|
.as_ref()
|
|
.unwrap()
|
|
.dispatch_error(webgpu::Error::Validation(String::from(
|
|
"Referenced GPUBuffer(s) are not Unmapped",
|
|
)));
|
|
return;
|
|
}
|
|
let command_buffers = command_buffers.iter().map(|cb| cb.id().0).collect();
|
|
self.channel
|
|
.0
|
|
.send(WebGPURequest::Submit {
|
|
queue_id: self.queue.0,
|
|
command_buffers,
|
|
})
|
|
.unwrap();
|
|
}
|
|
|
|
/// <https://gpuweb.github.io/gpuweb/#dom-gpuqueue-writebuffer>
|
|
#[allow(unsafe_code)]
|
|
fn WriteBuffer(
|
|
&self,
|
|
buffer: &GPUBuffer,
|
|
buffer_offset: GPUSize64,
|
|
data: BufferSource,
|
|
data_offset: GPUSize64,
|
|
size: Option<GPUSize64>,
|
|
) -> Fallible<()> {
|
|
let bytes = match data {
|
|
BufferSource::ArrayBufferView(d) => d.to_vec(),
|
|
BufferSource::ArrayBuffer(d) => d.to_vec(),
|
|
};
|
|
let content_size = if let Some(s) = size {
|
|
s
|
|
} else {
|
|
bytes.len() as GPUSize64 - data_offset
|
|
};
|
|
let valid = data_offset + content_size <= bytes.len() as u64 &&
|
|
buffer.state() == GPUBufferState::Unmapped &&
|
|
content_size % wgt::COPY_BUFFER_ALIGNMENT == 0 &&
|
|
buffer_offset % wgt::COPY_BUFFER_ALIGNMENT == 0;
|
|
|
|
if !valid {
|
|
return Err(Error::Operation);
|
|
}
|
|
|
|
let final_data = IpcSharedMemory::from_bytes(
|
|
&bytes[data_offset as usize..(data_offset + content_size) as usize],
|
|
);
|
|
if let Err(e) = self.channel.0.send(WebGPURequest::WriteBuffer {
|
|
queue_id: self.queue.0,
|
|
buffer_id: buffer.id().0,
|
|
buffer_offset,
|
|
data: final_data,
|
|
}) {
|
|
warn!("Failed to send WriteBuffer({:?}) ({})", buffer.id(), e);
|
|
return Err(Error::Operation);
|
|
}
|
|
|
|
Ok(())
|
|
}
|
|
|
|
/// <https://gpuweb.github.io/gpuweb/#dom-gpuqueue-writetexture>
|
|
fn WriteTexture(
|
|
&self,
|
|
destination: &GPUImageCopyTexture,
|
|
data: BufferSource,
|
|
data_layout: &GPUImageDataLayout,
|
|
size: GPUExtent3D,
|
|
) -> Fallible<()> {
|
|
let (bytes, len) = match data {
|
|
BufferSource::ArrayBufferView(d) => (d.to_vec(), d.len() as u64),
|
|
BufferSource::ArrayBuffer(d) => (d.to_vec(), d.len() as u64),
|
|
};
|
|
let valid = data_layout.offset <= len;
|
|
|
|
if !valid {
|
|
return Err(Error::Operation);
|
|
}
|
|
|
|
let texture_cv = convert_ic_texture(destination);
|
|
let texture_layout = convert_image_data_layout(data_layout);
|
|
let write_size = convert_texture_size_to_wgt(&convert_texture_size_to_dict(&size));
|
|
let final_data = IpcSharedMemory::from_bytes(&bytes);
|
|
|
|
if let Err(e) = self.channel.0.send(WebGPURequest::WriteTexture {
|
|
queue_id: self.queue.0,
|
|
texture_cv,
|
|
data_layout: texture_layout,
|
|
size: write_size,
|
|
data: final_data,
|
|
}) {
|
|
warn!(
|
|
"Failed to send WriteTexture({:?}) ({})",
|
|
destination.texture.id().0,
|
|
e
|
|
);
|
|
return Err(Error::Operation);
|
|
}
|
|
|
|
Ok(())
|
|
}
|
|
|
|
/// <https://gpuweb.github.io/gpuweb/#dom-gpuqueue-onsubmittedworkdone>
|
|
fn OnSubmittedWorkDone(&self) -> Rc<Promise> {
|
|
let global = self.global();
|
|
let promise = Promise::new(&global);
|
|
let sender = response_async(&promise, self);
|
|
if let Err(e) = self
|
|
.channel
|
|
.0
|
|
.send(WebGPURequest::QueueOnSubmittedWorkDone {
|
|
sender,
|
|
queue_id: self.queue.0,
|
|
})
|
|
{
|
|
warn!("QueueOnSubmittedWorkDone failed with {e}")
|
|
}
|
|
promise
|
|
}
|
|
}
|
|
|
|
impl AsyncWGPUListener for GPUQueue {
|
|
fn handle_response(
|
|
&self,
|
|
response: Option<Result<webgpu::WebGPUResponse, String>>,
|
|
promise: &Rc<Promise>,
|
|
) {
|
|
match response {
|
|
Some(Ok(WebGPUResponse::SubmittedWorkDone)) => {
|
|
promise.resolve_native(&());
|
|
},
|
|
_ => {
|
|
warn!("GPUQueue received wrong WebGPUResponse");
|
|
promise.reject_error(Error::Operation);
|
|
},
|
|
}
|
|
}
|
|
}
|