Struct Function 

pub struct Function { /* private fields */ }

A kernel entry point — either inside a Module (classic Driver API) or materialized from a crate::library::Kernel (CUDA 12.0+ library API). Either way it keeps the parent alive via an Arc so the kernel stays valid for as long as any Function handle exists.

Implementations

impl Function

pub fn launch(&self) -> LaunchBuilder<'_>

Start a kernel-launch builder for this function.

impl Function

pub fn as_raw(&self) -> CUfunction

Raw CUfunction. Use with care.

pub fn module(&self) -> Option<&Module>

The Module this kernel lives in, if it was obtained through Module::get_function. Returns None for kernels materialized from a library::Kernel.

pub fn get_attribute(&self, attribute: i32) -> Result<i32>

Query a kernel attribute (see baracuda_cuda_sys::types::CUfunction_attribute).

pub fn name(&self) -> Result<String>

Return the demangled kernel name reported by the driver.

pub fn param_info(&self, index: usize) -> Result<(usize, usize)>

Return (offset_in_bytes, size_in_bytes) for the index-th parameter in this function’s ABI signature.

pub fn module_raw(&self) -> Result<CUmodule>

Return the raw CUmodule this function was loaded from, if any.

pub fn set_attribute(&self, attribute: i32, value: i32) -> Result<()>

Set a kernel attribute. Only a subset is writable (notably MAX_DYNAMIC_SHARED_SIZE_BYTES and PREFERRED_SHARED_MEMORY_CARVEOUT).

pub fn max_threads_per_block(&self) -> Result<i32>

Maximum threads per block this kernel supports on the current device.

pub fn shared_size_bytes(&self) -> Result<i32>

Size of per-block statically-allocated shared memory (bytes).

pub fn num_regs(&self) -> Result<i32>

Number of registers used per thread.

pub fn local_size_bytes(&self) -> Result<i32>

Per-thread local-memory footprint (bytes).

pub fn ptx_version(&self) -> Result<i32>

PTX version this kernel was compiled from, as major*10 + minor.

pub fn binary_version(&self) -> Result<i32>

SM-architecture this kernel was compiled for, as major*10 + minor.

Trait Implementations

impl Clone for Function

fn clone(&self) -> Function

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for Function

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

Formats the value using the given formatter.

impl Send for Function

impl Sync for Function