About
Roofline Notes is a personal technical blog for processor architecture, GPU performance, algorithms, and LLM systems.
The site is built to support long-form writing, diagrams, short videos, interactive plots, and small tools that make hardware and performance ideas easier to inspect.
Topics
- Processor architecture and memory systems
- GPU kernels and performance analysis
- Roofline models and throughput limits
- LLM inference, batching, KV cache behavior, and quantization
- Algorithms as they meet real hardware constraints