0
0
Fork 0
mirror of https://github.com/denoland/deno.git synced 2024-10-29 08:58:01 -04:00
denoland-deno/ext/webgpu/queue.rs

161 lines
4.3 KiB
Rust
Raw Normal View History

// Copyright 2018-2021 the Deno authors. All rights reserved. MIT license.
use std::num::NonZeroU32;
use deno_core::error::bad_resource_id;
use deno_core::error::null_opbuf;
use deno_core::error::AnyError;
use deno_core::OpState;
use deno_core::ResourceId;
use deno_core::ZeroCopyBuf;
use serde::Deserialize;
use super::error::WebGpuResult;
2021-03-25 14:17:37 -04:00
type WebGpuQueue = super::WebGpuDevice;
#[derive(Deserialize)]
#[serde(rename_all = "camelCase")]
pub struct QueueSubmitArgs {
queue_rid: ResourceId,
command_buffers: Vec<u32>,
}
pub fn op_webgpu_queue_submit(
state: &mut OpState,
args: QueueSubmitArgs,
_: (),
) -> Result<WebGpuResult, AnyError> {
let instance = state.borrow::<super::Instance>();
let queue_resource = state
.resource_table
2021-03-25 14:17:37 -04:00
.get::<WebGpuQueue>(args.queue_rid)
.ok_or_else(bad_resource_id)?;
let queue = queue_resource.0;
let mut ids = vec![];
for rid in args.command_buffers {
let buffer_resource = state
.resource_table
2021-03-25 14:17:37 -04:00
.get::<super::command_encoder::WebGpuCommandBuffer>(rid)
.ok_or_else(bad_resource_id)?;
ids.push(buffer_resource.0);
}
let maybe_err =
gfx_select!(queue => instance.queue_submit(queue, &ids)).err();
Ok(WebGpuResult::maybe_err(maybe_err))
}
#[derive(Deserialize)]
#[serde(rename_all = "camelCase")]
2021-03-25 14:17:37 -04:00
struct GpuImageDataLayout {
offset: Option<u64>,
bytes_per_row: Option<u32>,
rows_per_image: Option<u32>,
}
#[derive(Deserialize)]
#[serde(rename_all = "camelCase")]
pub struct QueueWriteBufferArgs {
queue_rid: ResourceId,
buffer: u32,
buffer_offset: u64,
data_offset: usize,
size: Option<usize>,
}
pub fn op_webgpu_write_buffer(
state: &mut OpState,
args: QueueWriteBufferArgs,
zero_copy: Option<ZeroCopyBuf>,
) -> Result<WebGpuResult, AnyError> {
let zero_copy = zero_copy.ok_or_else(null_opbuf)?;
let instance = state.borrow::<super::Instance>();
let buffer_resource = state
.resource_table
2021-03-25 14:17:37 -04:00
.get::<super::buffer::WebGpuBuffer>(args.buffer)
.ok_or_else(bad_resource_id)?;
let buffer = buffer_resource.0;
let queue_resource = state
.resource_table
2021-03-25 14:17:37 -04:00
.get::<WebGpuQueue>(args.queue_rid)
.ok_or_else(bad_resource_id)?;
let queue = queue_resource.0;
let data = match args.size {
Some(size) => &zero_copy[args.data_offset..(args.data_offset + size)],
None => &zero_copy[args.data_offset..],
};
let maybe_err = gfx_select!(queue => instance.queue_write_buffer(
queue,
buffer,
args.buffer_offset,
data
))
.err();
Ok(WebGpuResult::maybe_err(maybe_err))
}
#[derive(Deserialize)]
#[serde(rename_all = "camelCase")]
pub struct QueueWriteTextureArgs {
queue_rid: ResourceId,
2021-03-25 14:17:37 -04:00
destination: super::command_encoder::GpuImageCopyTexture,
data_layout: GpuImageDataLayout,
size: super::texture::GpuExtent3D,
}
pub fn op_webgpu_write_texture(
state: &mut OpState,
args: QueueWriteTextureArgs,
zero_copy: Option<ZeroCopyBuf>,
) -> Result<WebGpuResult, AnyError> {
let zero_copy = zero_copy.ok_or_else(null_opbuf)?;
let instance = state.borrow::<super::Instance>();
let texture_resource = state
.resource_table
2021-03-25 14:17:37 -04:00
.get::<super::texture::WebGpuTexture>(args.destination.texture)
.ok_or_else(bad_resource_id)?;
let queue_resource = state
.resource_table
2021-03-25 14:17:37 -04:00
.get::<WebGpuQueue>(args.queue_rid)
.ok_or_else(bad_resource_id)?;
let queue = queue_resource.0;
let destination = wgpu_core::command::ImageCopyTexture {
texture: texture_resource.0,
mip_level: args.destination.mip_level.unwrap_or(0),
origin: args
.destination
.origin
.map_or(Default::default(), |origin| wgpu_types::Origin3d {
x: origin.x.unwrap_or(0),
y: origin.y.unwrap_or(0),
z: origin.z.unwrap_or(0),
}),
};
let data_layout = wgpu_types::ImageDataLayout {
offset: args.data_layout.offset.unwrap_or(0),
bytes_per_row: NonZeroU32::new(args.data_layout.bytes_per_row.unwrap_or(0)),
rows_per_image: NonZeroU32::new(
args.data_layout.rows_per_image.unwrap_or(0),
),
};
gfx_ok!(queue => instance.queue_write_texture(
queue,
&destination,
&*zero_copy,
&data_layout,
&wgpu_types::Extent3d {
width: args.size.width.unwrap_or(1),
height: args.size.height.unwrap_or(1),
depth_or_array_layers: args.size.depth_or_array_layers.unwrap_or(1),
}
))
}