Struct FftExecutionCache 

pub struct FftExecutionCache<'a> { /* private fields */ }

Execution-cache state supplied to an FftBackend.

Direct repeated calls use a caller-owned FftPlanCache, while traced execution uses the owning runtime’s [ExtensionCacheStore]. Both ownership paths expose the same bounded typed store to a backend, so CPU, Metal, CUDA, and future implementations can retain private plans or workspaces in their own cache namespace. Constructors keep the owner representation closed.

Examples

use tenferro_fft::{FftExecutionCache, FftPlanCache};

let mut plans = FftPlanCache::default();
let cache = FftExecutionCache::caller_owned(&mut plans);
assert!(format!("{cache:?}").contains("CallerOwned"));

Implementations

impl<'a> FftExecutionCache<'a>

pub fn caller_owned(cache: &'a mut FftPlanCache) -> Self

Build a context backed by a caller-owned typed FFT execution cache.

pub fn runtime_owned(cache: &'a mut ExtensionCacheStore) -> Self

Build a context backed by an extension runtime cache store.

pub fn store_mut(&mut self) -> &mut ExtensionCacheStore

Borrow the bounded typed store owned by the caller or extension runtime.

Backend implementations should use a stable family/cache namespace and include every plan-identity field in the key discriminator. The store owns LRU bounds, typed retrieval, clear behavior, entry counts, and the retained-byte estimates supplied at insertion.

Trait Implementations

impl Debug for FftExecutionCache<'_>

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

Formats the value using the given formatter.