Prefix-cache eviction: LRU by block or by request? And what does a radix tree change about it?
A prefix cache only hits from the first token forward, so evicting the wrong block can strand gigabytes that stay resident and can never be hit. The answer derives how much, and which eviction order guarantees it never happens.
Updated Oct 2026 · Worked answers on inference mechanisms, with calculations, implementation details, and follow-up questions.
A prefix cache only hits from the first token forward, so evicting the wrong block can strand gigabytes that stay resident and can never be hit. The answer derives how much, and which eviction order guarantees it never happens.
At batch 1 the step reads every weight once to produce one token per sequence, so arithmetic intensity sits near 1 FLOP per byte and the tensor cores idle.