inference/research
drafthigh · 0.85updated 2026-10-03

LMCache GitHub repository

repo
This record is a draft. Not yet verified. Provisional. Treat every number on this page as provisional.

Fields

id
sources/kern-lmcache-github
Publisher
LMCache
Kind
repo
Accessed
2026-10-03
notes
README, Apache-2.0. This is a KV-cache movement layer, not a GPU kernel library, so it is recorded as a source for the claim that KV traffic (and therefore the attention kernel's HBM read volume) is set by a cache tier decision made outside the engine. No performance numbers taken from it here.

Notes

README, Apache-2.0. This is a KV-cache movement layer, not a GPU kernel library, so it is recorded as a source for the claim that KV traffic (and therefore the attention kernel's HBM read volume) is set by a cache tier decision made outside the engine. No performance numbers taken from it here.

Cross-references

This record references no other records.