Module wgpu_core::hub

source ·
Expand description

Allocating resource ids, and tracking the resources they refer to.

The wgpu_core API uses identifiers of type Id<R> to refer to resources of type R. For example, id::DeviceId is an alias for Id<Device<Empty>>, and id::BufferId is an alias for Id<Buffer<Empty>>. Id implements Copy, Hash, Eq, Ord, and of course Debug.

Each Id contains not only an index for the resource it denotes but also a Backend indicating which wgpu backend it belongs to. You can use the gfx_select macro to dynamically dispatch on an id’s backend to a function specialized at compile time for a specific backend. See that macro’s documentation for details.

Ids also incorporate a generation number, for additional validation.

The resources to which identifiers refer are freed explicitly. Attempting to use an identifier for a resource that has been freed elicits an error result.

Assigning ids to resources

The users of wgpu_core generally want resource ids to be assigned in one of two ways:

  • Users like wgpu want wgpu_core to assign ids to resources itself. For example, wgpu expects to call Global::device_create_buffer and have the return value indicate the newly created buffer’s id.

  • Users like player and Firefox want to allocate ids themselves, and pass Global::device_create_buffer and friends the id to assign the new resource.

To accommodate either pattern, wgpu_core methods that create resources all expect an id_in argument that the caller can use to specify the id, and they all return the id used. For example, the declaration of Global::device_create_buffer looks like this:

impl<G: GlobalIdentityHandlerFactory> Global<G> {
    /* ... */
    pub fn device_create_buffer<A: HalApi>(
        &self,
        device_id: id::DeviceId,
        desc: &resource::BufferDescriptor,
        id_in: Input<G, id::BufferId>,
    ) -> (id::BufferId, Option<resource::CreateBufferError>) {
        /* ... */
    }
    /* ... */
}

Users that want to assign resource ids themselves pass in the id they want as the id_in argument, whereas users that want wgpu_core itself to choose ids always pass (). In either case, the id ultimately assigned is returned as the first element of the tuple.

Producing true identifiers from id_in values is the job of an crate::identity::IdentityManager, but only if the IdentityHandlerFactory create it and then generated by it, otherwise ids will be received from outside.

Global::new expects a factory argument that implements the GlobalIdentityHandlerFactory trait, which extends crate::identity::IdentityHandlerFactory<I> for each resource id type I. This trait, in turn, has a spawn method that constructs an crate::identity::IdentityManager for the Global to use, if ids should be generated by wgpu or will return None otherwise.

Id allocation and streaming

Perhaps surprisingly, allowing users to assign resource ids themselves enables major performance improvements in some applications.

The wgpu_core API is designed for use by Firefox’s WebGPU implementation. For security, web content and GPU use must be kept segregated in separate processes, with all interaction between them mediated by an inter-process communication protocol. As web content uses the WebGPU API, the content process sends messages to the GPU process, which interacts with the platform’s GPU APIs on content’s behalf, occasionally sending results back.

In a classic Rust API, a resource allocation function takes parameters describing the resource to create, and if creation succeeds, it returns the resource id in a Result::Ok value. However, this design is a poor fit for the split-process design described above: content must wait for the reply to its buffer-creation message (say) before it can know which id it can use in the next message that uses that buffer. On a common usage pattern, the classic Rust design imposes the latency of a full cross-process round trip.

We can avoid incurring these round-trip latencies simply by letting the content process assign resource ids itself. With this approach, content can choose an id for the new buffer, send a message to create the buffer, and then immediately send the next message operating on that buffer, since it already knows its id. Allowing content and GPU process activity to be pipelined greatly improves throughput.

To help propagate errors correctly in this style of usage, when resource creation fails, the id supplied for that resource is marked to indicate as much, allowing subsequent operations using that id to be properly flagged as errors as well.

Structs