Module memory 

Structs

AllocatorMemorySnapshot

AllocatorPressureRelief

MemoryEstimate

A cold-path estimate of memory AFT can attribute without allocator hooks.

MemorySnapshot

ProcessMemorySnapshot

RootMemorySnapshot

SqliteMemorySnapshot

Enums

AllocatorSource

Constants

ALLOCATOR_SLACK_RELIEF_MIN_INTERVAL

Minimum spacing between opportunistic relief passes so a workload that legitimately cycles through large allocations does not thrash the allocator.

ALLOCATOR_SLACK_RELIEF_THRESHOLD_BYTES

Allocator slack (mapped-but-unused arena bytes) above which opportunistic pressure relief is worth the zone-lock contention it briefly causes.

MEMORY_SNAPSHOT_ROOT_DETAIL_CAP

Cap on per-root detail entries in serialized snapshots. Process totals always cover every root; only the per-root breakdown is capped so a many-root daemon process cannot balloon the status payload past downstream consumers’ size limits (the daemon metrics cache truncates around 27 KB, and JSON keys serialize alphabetically, so an oversized memory.roots map pushes later sections past the cut).

Functions

allocator_slack_relief_due

Decide whether an opportunistic allocator relief pass is due.

cached_allocator_observation

Return the last process-wide allocator observation without walking allocator zones. The observation timestamp is separate from the sampler reservation so a status request never mistakes an in-progress sample for a fresh reading.

estimated_json_bytes

last_allocator_relief_at_ms

last_allocator_relief_freed_bytes

path_bytes

relieve_allocator_pressure

spawn_allocator_slack_relief_if_due

Opportunistically return unused allocator pages without reading allocator statistics on the module loop. The loop only reads cached atomics. When the cached observation is stale, one sampler takes an initial snapshot; if it finds enough slack, the page-return operation reuses it and takes only one additional after snapshot, limiting the pass to two statistics walks.

usize_to_u64