Struct TensorMapArg

pub struct TensorMapArg<'a, R: Runtime> {
    pub tensor: TensorArg<'a, R>,
    pub metadata: TensorMapMeta,
}

Grid constant tensor map, currently only maps to CUDA tensormap. May be interleaved or swizzled, but last dimension must be contiguous (since strides don’t include the last dimension).

The tensormap is treated as an opaque type at runtime.

Fields

tensor: TensorArg<'a, R>

metadata: TensorMapMeta

Implementations

impl<'a, R: Runtime> TensorMapArg<'a, R>

pub fn new( format: TensorMapFormat, tensor: TensorArg<'a, R>, elem: Elem, ) -> Self

pub fn with_elem_stride(self, elem_stride: Vec<usize>) -> Self

pub fn with_interleave(self, interleave: TensorMapInterleave) -> Self

pub fn with_swizzle(self, swizzle: TensorMapSwizzle) -> Self

pub fn with_prefetch(self, prefetch: TensorMapPrefetch) -> Self

pub fn with_nan_fill(self) -> Self

Trait Implementations

impl<R: Runtime> ArgSettings<R> for TensorMapArg<'_, R>

fn register(&self, launcher: &mut KernelLauncher<R>)

Register the information of an argument to the KernelLauncher.