Multi Level Cache that Does Up To A Billion Lookups Per Second in C++

Caching algorithm that has 1-2 nanosecond average latency to each cache-hit and not-too-slow cache-miss functions to the middle layer LRU. CLOCK-second-chance algorithm is used in this middle layer cache implementation and it is fast because of not-allocating any new node during cache eviction. For multithreading & read-only access, a third synchronized LRU is used.