Struct WgpuRuntime 

pub struct WgpuRuntime { /* private fields */ }

Headless wgpu runtime for compute-only workloads.

Implementations

impl WgpuRuntime

pub fn new_headless() -> SpatialResult<Self>

Creates a headless wgpu runtime preferring a high-performance adapter.

Prefer Self::shared when running multiple GPU filters in one process.

pub fn new_headless_with_preference( preference: WgpuPowerPreference, ) -> SpatialResult<Self>

Creates a headless wgpu runtime with an explicit adapter power preference.

pub fn shared() -> SpatialResult<Arc<Self>>

Returns a process-wide shared headless runtime, initializing it on first use.

pub fn device(&self) -> &Device

Returns the underlying wgpu device.

pub fn queue(&self) -> &Queue

Returns the underlying wgpu queue.

pub const fn adapter_info(&self) -> &WgpuAdapterInfo

Returns the selected adapter/backend receipt.

pub fn wait_idle(&self)

Blocks until all previously submitted work on this device is complete.

Normal device-resident chains do not synchronize implicitly. This is an explicit profiling/testing boundary or host-coordination primitive.

pub fn cached_image_texture_count(&self) -> usize

Returns the number of image textures retained for steady-state reuse.

pub fn initialized_image_pipeline_count(&self) -> usize

Returns the number of initialized GPU image pipeline families.

pub fn pipelines(&self) -> &ComputePipelineCache

Returns cached compute pipelines for this runtime’s device.

pub fn max_gather_channels(&self) -> u32

Returns the maximum attribute channels gatherable in one multi dispatch.

pub fn buffer_pool(&self) -> &GpuBufferPool

Returns the reusable storage-buffer pool owned by this runtime.

pub fn upload_pod_storage<T: Pod>( &self, label: &'static str, data: &[T], ) -> SpatialResult<Buffer>

Uploads a POD slice into a reusable pooled storage buffer.

pub fn upload_f32_storage( &self, label: &'static str, data: &[f32], ) -> SpatialResult<Buffer>

Uploads f32 values into a reusable pooled storage buffer.

pub fn upload_u32_storage( &self, label: &'static str, data: &[u32], ) -> SpatialResult<Buffer>

Uploads u32 values into a reusable pooled storage buffer.

pub fn recycle_storage(&self, byte_len: u64, buffer: Buffer)

Returns a storage buffer to the upload pool for reuse.

pub fn clear_buffer_pool(&self)

Clears all buffers cached in the upload pool.

pub async fn new_headless_async( preference: WgpuPowerPreference, ) -> SpatialResult<Self>

Creates a headless runtime asynchronously.

This is the browser/WebGPU construction path because blocking a WASM main thread is not supported. Native callers may also use it from their async executor.

Trait Implementations

impl SpatialRuntime for WgpuRuntime

Available on crate feature gpu-wgpu only.

fn device_kind(&self) -> DeviceKind

Returns the device kind owned by this runtime.

fn supports_policy(&self, policy: ExecutionPolicy) -> bool

Returns whether this runtime can satisfy the requested policy.