# oMLX ProcessMemoryEnforcer pressure levels and adjust_store_cache_cap > Pressure levels are ok below the soft mark, soft between soft and hard, and hard at or above the hard mark, both marks being fractions of the recomputed hard-limit ceiling; the ceiling moves every tick with system availability. Parent: [Mac local LLMs: oMLX, Rapid-MLX and related internals](https://llms-explorer.com/tree/mac-local-llms-omlx-and-rapid-mlx-internals/) · 2 facets · 16 facts · page: https://llms-explorer.com/tree/omlx-processmemoryenforcer-pressure-levels-and-a/ ## Facts - Pressure levels are ok below the soft mark, soft between soft and hard, and hard at or above the hard mark, both marks being fractions of the recomputed hard-limit ceiling; the ceiling moves every tick with system availability. — [source](https://raw.githubusercontent.com/jundot/omlx/main/omlx/process_memory_enforcer.py) - At soft pressure the enforcer evicts idle non-pinned LRU models only while more than one is loaded, pauses new admissions, and leaves in-flight requests alone. — [source](https://raw.githubusercontent.com/jundot/omlx/main/omlx/process_memory_enforcer.py) - At hard pressure it unloads even the last idle non-pinned model, aborting its requests first, since an idle model holds no KV worth keeping. — [source](https://raw.githubusercontent.com/jundot/omlx/main/omlx/process_memory_enforcer.py) - At hard pressure with a sole busy non-pinned victim it aborts that model's requests and keeps the model loaded unless emergency pressure holds, in which case it marks a pending unload. — [source](https://raw.githubusercontent.com/jundot/omlx/main/omlx/process_memory_enforcer.py) - At hard pressure with nothing evictable it first flags in-progress model loads to abort, and only if none exist does it abort active requests (emergency only) or call `request_idle_reclaim` on each scheduler. — [source](https://raw.githubusercontent.com/jundot/omlx/main/omlx/process_memory_enforcer.py) - Emergency pressure is not the hard watermark: it requires usage at or above the real ceiling and either 2 GiB over it or 2 consecutive polls over it (`_EMERGENCY_OVER_CEILING_MARGIN_BYTES`, `_EMERGENCY_OVER_CEILING_POLLS`). — [source](https://raw.githubusercontent.com/jundot/omlx/main/omlx/process_memory_enforcer.py) - Before pausing admission or evicting, a tick at or above soft calls `clear_image_decode_cache` and re-measures usage. — [source](https://raw.githubusercontent.com/jundot/omlx/main/omlx/process_memory_enforcer.py) - When the level is not ok and not emergency, `request_pressure_reclaim` is sent to every resolvable scheduler and enforcement is deferred for at most `_PRESSURE_RECLAIM_GRACE_POLLS_MAX = 5` consecutive polls per pressure episode; the counter resets when the level returns to ok. — [source](https://raw.githubusercontent.com/jundot/omlx/main/omlx/process_memory_enforcer.py) - The reclaim request is skipped unless the MLX buffer pool exceeds `_POOL_RECLAIM_FLOOR` (2 GiB) or a hot-cache shrink just freed references. — [source](https://raw.githubusercontent.com/jundot/omlx/main/omlx/process_memory_enforcer.py) - The enforcer never touches Metal itself: it sets a GIL-atomic flag and the scheduler clears the cache on the inference thread at the next step boundary, even under load, unlike `request_idle_reclaim` which waits for idle. — [source](https://raw.githubusercontent.com/jundot/omlx/main/omlx/process_memory_enforcer.py) - After a hot-cache shrink at hard pressure the level is recomputed, and the enforcer can drop from hard to soft or ok without evicting a model. — [source](https://raw.githubusercontent.com/jundot/omlx/main/omlx/process_memory_enforcer.py) - `_walk_store_cache_caps` calls `scheduler.adjust_store_cache_cap(pressure_level)` on every tick, including deferred ticks and ok ticks, so the cap moves at most one step per poll; at ok it recovers toward `max_num_seqs`. — [source](https://raw.githubusercontent.com/jundot/omlx/main/omlx/process_memory_enforcer.py) - Poll interval is 1.0 s while pressure is not ok, an activity hint is live, a model is loading, or any engine has active requests (or cannot report them); 10 s when models are loaded and idle; 30 s when none are loaded. — [source](https://raw.githubusercontent.com/jundot/omlx/main/omlx/process_memory_enforcer.py) - `wake()` sets an event so the loop re-checks early and extends an activity hint of at least 2 s. — [source](https://raw.githubusercontent.com/jundot/omlx/main/omlx/process_memory_enforcer.py) - The adaptive prefill chunk sizer uses fractions of the hard ceiling of 0.90 safe, 0.92 balanced, 0.97 aggressive and 0.95 custom; the pre-chunk abort margin uses 0.90, 0.93, 0.97 and 0.95. — [source](https://raw.githubusercontent.com/jundot/omlx/main/omlx/process_memory_enforcer.py) ## Corrections and disagreements - CONTRADICTS: omlx-store-cache-admission-gate-and-60-s-admissi.md only in timing granularity: the cap walk runs per enforcement tick (1 s under pressure, but 10 s or 30 s when idle and ok), not per pressure transition. — [source](https://raw.githubusercontent.com/jundot/omlx/main/omlx/process_memory_enforcer.py)