Everything this blog can render
A reference post that exercises every piece of Markdown syntax the site supports — headings, code, math, callouts, tables, footnotes, images and the awkward edge cases — so regressions are obvious.
Writing
11 posts, newest first.
聊聊 KV Cache 的显存账：一张卡到底能塞多少并发
把 KV Cache 的显存占用算清楚，再看 PagedAttention 的分块管理省下的是哪一部分。含一份可以直接抄去改的估算脚本，以及几个在生产里反复踩到的坑。
Continuous batching, and where the throughput actually goes
A scheduler-level look at continuous batching: why naive static batching wastes most of the GPU, what the arrival-rate math says about queue depth, and the three places real implementations lose the throughput they just won.