Struct PrecompileCache 

pub struct PrecompileCache { /* private fields */ }

Per-block cache for precompile results shared between warmer and executor.

Holds at most its budget, PRECOMPILE_CACHE_MAX_BYTES unless built with PrecompileCache::with_max_bytes, as counted by PrecompileCache::entry_size; once full, further results are simply not cached. Nothing is evicted: the cache is dropped with its block, so there is no recency worth tracking, and lookups stay on the shared read lock.

Implementations

impl PrecompileCache

pub fn new() -> Self

pub fn with_max_bytes(max_bytes: usize) -> Self

A cache that holds at most max_bytes, as counted by PrecompileCache::entry_size.

pub fn entry_size(calldata_len: usize, output_len: usize) -> usize

Bytes one entry is charged against the budget: its calldata and output, plus the fixed size of the key and value it is stored as. Hash-table slack and the per-precompile tables are not counted, so the budget is approximate, not an exact resident size.

pub fn get(&self, address: &Address, calldata: &[u8]) -> Option<(Bytes, u64)>

pub fn insert( &self, address: Address, calldata: &[u8], output: &[u8], gas_cost: u64, )

Stores copies of calldata and output, never the caller’s buffers: calldata can be a view of a larger allocation (the calling frame’s memory, or a slice of a transaction’s data), and keeping a reference would hold that whole allocation while the budget counts only the length.

The copies are made only for a call that is not cached yet and fits the budget, and outside the write lock. The warmer and the executor often compute the same call, and the second insert then costs a lookup under the shared read lock and nothing more.

Trait Implementations

impl Default for PrecompileCache

fn default() -> Self

Returns the “default value” for a type.