Module arena 

Arena allocator for batch operations

This module provides arena-based memory allocation to reduce allocation overhead in hot paths and improve memory locality.

Features (ADR-001)

• Cache-aligned allocations: All allocations are aligned to cache line boundaries (64 bytes)
• Bump allocation: O(1) allocation with minimal overhead
• Batch deallocation: Free all allocations at once via reset()
• Thread-local arenas: Per-thread allocation without synchronization

Structs

Arena

Arena allocator for temporary allocations

ArenaVec

Vector allocated from an arena

BatchVectorAllocator

Batch vector allocator for processing multiple vectors (ADR-001)

CacheAlignedVec

Get the thread-local arena Note: Commented out due to lifetime issues with RefCell::borrow() escaping closure Use THREAD_ARENA.with(|arena| { … }) directly instead Cache-aligned vector storage for SIMD operations (ADR-001)

Constants

CACHE_LINE_SIZE

Cache line size (typically 64 bytes on modern CPUs)