Struct LocalBuffer

pub struct LocalBuffer<T> { /* private fields */ }

A buffer on the GPU.

The concept of a local buffer is from OpenCL. In CUDA you don’t allocate a buffer directly via API call. Instead you pass in the amount of shared memory that should be used.

There can be at most a single local buffer per kernel. On CUDA a null pointer will be passed in, instead of an actual value. The memory that should get allocated is then passed into the kernel call automatically.

Implementations

impl<T> LocalBuffer<T>

pub fn new(length: usize) -> Self

Returns a new buffer of the specified length.

Trait Implementations

impl<T: Debug> Debug for LocalBuffer<T>

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

Formats the value using the given formatter.

impl<T> KernelArgument for LocalBuffer<T>

fn as_c_void(&self) -> *mut c_void

Converts into a C void pointer.

fn shared_mem(&self) -> u32

Returns the shared memory size. This is usally 0, except for LocalBuffers. This informations is used to allocate the memory correctly.

impl<T> KernelArgument for LocalBuffer<T>

fn push(&self, kernel: &mut Kernel<'_>)

Apply the kernel argument to the kernel.