use ;
/// Backend-selected per-channel 3-vector cross product.
///
/// Standalone capability, not a [`BackendOps`](super::super::BackendOps)
/// bundle member: every backend that exists today already computes
/// `coeus_ops::cross` through the shared host-fold path
/// (`crate::backend_ops::defaults::cross::cross_fold`), so folding this
/// into the required bundle would force an identical no-op impl onto every
/// backend for a capability `dot`/`cross` never required before. A backend
/// with an on-device seam (currently `WgpuBackend`, for the contiguous
/// last-axis layout) overrides [`CrossOps::cross_storage`] instead of
/// relying on the shared default.