Starting a blog
I spend most of my days making GPUs faster — chasing kernel stalls, reading occupancy counters, and arguing with quantization error. A lot of what I learn never makes it into a paper: it's too small, too practical, or too tied to one chip. This blog is where those notes go.
Expect short, concrete posts on GPU computing, AI inference on mobile and HPC hardware, computer architecture, and the occasional medical-imaging tangent. Some will be in English, some in 中文 — each post says which.
Why a static blog
This site is three files and no build step, and I'd like to keep it that way. Each post is a plain HTML file; the archive is a hand-maintained list. No framework to upgrade, nothing to break in two years. If you can open a text editor, you can read the source of every post here.
More soon.