# Memory

allocation patterns, leaks, heap pressure, fragmentation.

## Measure

- heap profiling → allocation rate, live heap, pause/freq of reclamation
- allocation tracking → hot allocators, sites
- reclamation logs → frequency, duration, cause
- metrics → bytes/sec, heap trend, reclamation overhead %

## Optimize

- reuse objects/buffers for short-lived values
- avoid allocations in hot loops (pre-allocate, reuse buffers)
- efficient data structures (contiguous over pointer-chasing for sequential access)
- tune allocator/reclamation params if applicable
- size objects to fit cache lines; avoid fragmentation
- avoid boxing/copying where the runtime does it implicitly

## Pitfalls

- premature opt: small alloc reductions may not matter if allocator is fast
- over-reusing: complexity + hidden leaks
- ignoring reclamation: reducing allocs may not help if reclamation already efficient
- peak heap ≠ average; average matters for throughput