Struct FillPlan 

pub struct FillPlan<T: Element> { /* private fields */ }

Fill plan.

y[i] = value for all i (PyTorch torch.full).

When to use: zero-init / constant-init of output buffers, or any broadcast-fill. No BW — a constant tensor has zero gradient.

Dtypes: {f32, f64, f16, bf16, i32, i64} — every numeric Element dtype baracuda exposes through the unified Plan layer. u8 / i8 ship in the sys crate but on the IntElement family (deferred plan shape).

Shape limits: flat [numel]; numel ≥ 0.

Workspace: none.

Precision guarantee: deterministic, bit-stable, bit-exact. f16 / bf16 transport value via raw u16 bit pattern; the safe plan layer performs the bit cast.

Implementations

impl<T: Element> FillPlan<T>

pub fn select( _stream: &Stream, desc: &FillDescriptor<T>, _pref: PlanPreference, ) -> Result<Self>

Pick a kernel for desc.

pub fn can_implement(&self, args: &FillArgs<'_, T>) -> Result<()>

Validate args.

pub fn workspace_size(&self) -> usize

Workspace size in bytes. Always 0.

pub fn sku(&self) -> KernelSku

Identity of the kernel this plan picked.

pub fn precision_guarantee(&self) -> PrecisionGuarantee

Numerical guarantees for this plan’s kernel.

pub fn run( &self, stream: &Stream, _workspace: Workspace<'_>, args: FillArgs<'_, T>, ) -> Result<()>

Launch.