ben-manes / caffeine / #5684

100%

master: 100%

push

github

Adapt the window by hit density with starvation-guarded probes

Large caches size the admission window by the within-sample hit density
of the two regions, ln(windowDensity / mainDensity), taking proportional
steps that are immune to the cross-sample hit rate swings of a phasey
workload. The signal is resident-only, so a region earning ~nothing is
its blind state: it cannot see what a different split would earn, and
holding such a position can pin the window at an extreme (constructible
steady-state workloads pinned it ~28pp below LRU at any scale).

Starved samples at a blind corner therefore probe by hit rate. Walks
are bounded by a sample budget, by a reversal that crosses back through
the probe's start, and by an absolute hit rate veto that undoes damage;
the density signal adjudicates each endpoint under strict confirmation.
A crash re-arms the refractory without doubling, since an exogenous
workload shift is indistinguishable from probe damage and mispricing it
would starve the re-exploration the shift calls for. Escalating
commitment resolves the stray-exit dilemma: first-round probes exit
cheaply when the watched region earns without density confirmation
(stray and transferred hits scale with a region's size, and thin-signal
workloads correctly at the floor must not pay deep-walk displacement),
while each adjudicated failure lengthens the refractory ladder and its
deeper rungs commit the walk past the stray zone, converting a deep
silent reuse band from an absorbing pin into a bounded, temporary dip.
Deep rungs also stride at 2x and 4x (capped at the 30% max step): the
ladder previously bought a committed walk permission to pass the stray
zone but not speed, so deep reuse bands were reached only by outlasting
the budget, and stray walls calibrated wider than a flat stride's reach
were absorbing. Failed walks still undo in full, which is what
distinguishes this from the rejected travel-budget exits. The sample
period is capped by the sketch's e... (continued)

4305 of 4315 branches covered (99.77%)

196 of 196 new or added lines in 6 files covered. (100.0%)

8601 of 8602 relevant lines covered (99.99%)

1.0 hits per line