Struct Ctx 

pub struct Ctx {
    pub tier: DeviceTier,
    pub device: Device,
    pub queue: Queue,
    pub capabilities: DeviceCapabilities,
}

A ready-to-use GPU context.

Fields

tier: DeviceTier

The tier the device was requested at.

Kept beside the device because a device reports what it was given, not what was asked for: a WebGL2-shaped device on a desktop adapter has the adapter’s features and limits as lowered by the tier, which is only readable as a description when the tier is part of it.

device: Device

The device every resource in a test is created on.

queue: Queue

The queue a test’s writes and submissions go through.

capabilities: DeviceCapabilities

What the device can do beyond the WebGPU baseline.

A DeviceCapabilities rather than the adapter it was read from: the adapter is dropped once the device exists, and the renderer needs the capabilities, not the adapter.

Implementations

impl Ctx

pub async fn headless() -> Ctx

Create a headless GPU context with the default adapter.

UNLIT3D_DEVICE_TIER narrows the device; see Ctx::headless_for.

Asynchronous because wgpu’s adapter and device requests are: on the web they wrap promises that only settle in a later task, and a browser has no second thread to block on. A synchronous caller drives this through block_on.

pub async fn headless_for(tier: DeviceTier) -> Ctx

Create a headless GPU context restricted to tier.

The tier’s limits are requested from the adapter and its capabilities are recorded beside the device, so a test can exercise the paths a WebGL2 browser takes — no storage buffers, no base_vertex — on hardware that has both. The adapter’s own limits and capabilities are what DeviceTier::Native asks for; the WebGPU baseline’s, which is what Ctx::headless uses, are what DeviceTier::WebGpu asks for.

Nothing is presented: the context is headless on the web too, where the canvas exists only because a browser will not hand out an adapter without one.