Module optimizations 

Internal performance optimizations for the native backend.

This module contains various in-memory optimizations that accelerate graph traversal without changing the on-disk format or public APIs. All optimizations are internal implementation details only.

INLINE HINT STRATEGY (Phase 13 Step 5)

Tier A: Tiny Hot Path Functions (#[inline(always)])

• unlikely(): Simple boolean wrapper used frequently in hot paths
• get_outgoing_edges(), get_incoming_edges(): Simple HashMap access
• get(), put(): Simple cache operations
• Thread-local access wrappers: Minimal overhead functions

Tier B: Small Helper Functions (#[inline] or compiler-driven)

• register_edge(): Moderate complexity but hot path
• Cache management functions: Used frequently but larger logic

Tier C: Large Functions (no inline hints)

• Cache eviction logic: Complex algorithms left to compiler discretion

Structs

NeighborPointerTable

Neighbor pointer table mapping node IDs to lists of edge file offsets This enables O(1) adjacency lookup without scanning edge slots

NodeHot

Hot-field node cache containing frequently accessed node metadata This avoids full node deserialization for adjacency operations

NodeHotCache

Functions

clear_all_caches

Clear all optimization caches (useful for testing)

extract_node_hot

Extract hot metadata from a node record

get_incoming_edge_offsets

Get incoming edge offsets for a node from the pointer table

get_node_hot

Get hot metadata for a node from the cache

get_outgoing_edge_offsets

Get outgoing edge offsets for a node from the pointer table

put_node_hot

Put hot metadata for a node in the cache

register_edge_offset

Register an edge in the neighbor pointer table This should be called whenever an edge is written to disk Optimized with cold-path offloading for cache management

with_neighbor_pointer_table

Get access to the global neighbor pointer table

with_node_hot_cache

Get access to the global node hot cache