Notes

Dumping all my notes here: what I write down while studying a subject, published as I go.

What's here

• C++: the language from the ground up: object lifetime and value semantics, the type system, how a program is built and run, the STL, and concurrency down to memory order.
• Performance: memory and object pools, thread pools, ring buffers, CPU caches and timing wheels, written with low-latency systems in mind.
• Operating systems: a four-week pass through virtual memory, traps, scheduling and I/O, then MIT 6.1810: one note per xv6 lab and the Unix fundamentals underneath them.
• Networking: networking fundamentals and gRPC.
• LLM: how a language model is trained and run, written alongside mini-LLM, a small decoder-only model in C++ from scratch: the math, attention shape by shape, training, inference and the KV cache, then parallelism, vLLM and the software stack underneath.
• Infrastructure: how Docker, Kubernetes and Terraform layer, and how mini-infra deploys a small LLM-serving app with each of them.

Most notes are written in Chinese.