Struct AsyncKernel 

pub struct AsyncKernel { /* private fields */ }

A kernel wrapper with async launch capability.

Wraps a Kernel and provides methods that return Futures resolving when the GPU work completes.

Implementations

impl AsyncKernel

pub fn new(kernel: Kernel) -> Self

Creates a new AsyncKernel with default configuration.

pub fn with_config(kernel: Kernel, config: AsyncLaunchConfig) -> Self

Creates a new AsyncKernel with the given configuration.

pub fn kernel(&self) -> &Kernel

Returns a reference to the underlying Kernel.

pub fn name(&self) -> &str

Returns the kernel function name.

pub fn config(&self) -> &AsyncLaunchConfig

Returns a reference to the current AsyncLaunchConfig.

pub fn set_config(&mut self, config: AsyncLaunchConfig)

Updates the async configuration.

pub fn launch_async<A: KernelArgs>( &self, params: &LaunchParams, stream: &Stream, args: &A, ) -> CudaResult<LaunchCompletion>

Launches the kernel and returns a LaunchCompletion future.

The kernel is launched asynchronously on the given stream, then a CUDA event is recorded. The returned future polls that event until it completes.

Errors

Returns a CudaError if the kernel launch or event operations fail. The future itself can also resolve to an error if the event query fails later.

pub fn launch_and_time_async<A: KernelArgs>( &self, params: &LaunchParams, stream: &Stream, args: &A, ) -> CudaResult<TimedLaunchCompletion>

Launches the kernel and returns a TimedLaunchCompletion future that resolves to LaunchTiming with elapsed GPU time.

Two events are recorded: one before and one after the kernel launch. When the future resolves, the elapsed time between the two events is computed.

Errors

Returns a CudaError if the launch or event operations fail.

Trait Implementations

impl Debug for AsyncKernel

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Display for AsyncKernel

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.