Struct LayeredPipelineCache 

pub struct LayeredPipelineCache { /* private fields */ }

Composite store that reads from every backend and writes to the first. Lets callers compose [RamStore, DiskStore, RemoteStore] so a miss at the fast layer falls through to slower layers.

Implementations

impl LayeredPipelineCache

pub fn new(layers: Vec<Arc<dyn PipelineCacheStore>>) -> Self

Construct from an ordered list (fastest-first). Lookups consult every layer in order; writes land in the first layer only - downstream layers are expected to be populated independently (e.g., from a pre-compiled blob bundle).

pub fn promotion_report(&self) -> LayeredPromotionReport

Snapshot promotion evidence for lower-layer hits copied into faster preceding layers.

Trait Implementations

impl PipelineCacheStore for LayeredPipelineCache

fn get_arc(&self, fp: &PipelineFingerprint) -> Option<Arc<Vec<u8>>>

V7-PERF-009: forward through to each layer’s zero-clone path so the hit propagates without an intermediate Vec<u8> allocation.

fn get(&self, fp: &PipelineFingerprint) -> Option<Vec<u8>>

Look up a cached artifact for this fingerprint.

fn put(&self, fp: PipelineFingerprint, artifact: Vec<u8>)

Insert a pre-compiled artifact. Implementations may dedupe or evict per their own policy.

fn flush(&self) -> Result<()>

Durably flush artifacts that were accepted by Self::put.

fn metrics(&self) -> PipelineCacheMetrics

Snapshot cache instrumentation for latency, throughput, and eviction gates. Backends that do not maintain counters return zeros.