|
|
template<class Word> |
| constexpr Word | pixie::experimental::slot_order_detail::mask (std::size_t bits) noexcept |
| |
|
template<class Word, std::size_t Words> |
| void | pixie::experimental::slot_order_detail::rotate_bits (std::array< Word, Words > &words, std::size_t left, std::size_t length, std::size_t distance) noexcept |
| |
|
template<std::size_t N> |
| void | pixie::experimental::slot_order_detail::rotate_bytes (std::array< std::uint8_t, N > &bytes, std::size_t left, std::size_t right, std::size_t distance) noexcept |
| |
Experimental split-bit and byte orders for 32/64 physical slots.
Split order encoding occupies 20/48 bytes; complete objects occupy 24/56 bytes, versus 33/65 for byte orders. Physical occupancy is dense, so count represents it without a stored bitmap. Split-bit rotation is scalar; byte rotation dispatches locally to supported SIMD or a scalar fallback. Primitive probes exclude tree traversal, subtree measures, and allocation. Current primitive and integrated measurements are retained in benchmarks/sequence_snapshot.md.