Learn Kernels
Curated by Learn Kernels • Added 2026-07-10
Interactive book on GPU kernels and inference performance traced to primary sources
Learn Kernels provides a first-principles breakdown of GPU computing architecture, memory hierarchies, CUDA threads, warp scheduling, coalesced access patterns, and writing high-performance Triton kernels for AI workloads.
What You Will Learn
- ✓GPU memory architecture: registers, shared memory (SRAM), L2 cache, and HBM throughput
- ✓Warp execution mechanics, divergence penalties, and coalesced global memory loads
- ✓Implementing and optimizing General Matrix Multiplication (GEMM) from scratch
- ✓Authoring custom fused attention kernels using OpenAI Triton and benchmarking against cuBLAS
Curator Highlights
- ★Interactive memory hierarchy bandwidth diagrams
- ★Direct primary source paper traces
- ★Rigorous systems engineering depth
Related in AI & Machine Learning
Other vetted resources in this discipline
TrenTorch
AI & Machine Learning•Interactive
Build machine learning frameworks from scratch by hand
Fanout
AI & Machine Learning•Platform
Mapped technical learning across AI research, systems, and ML math
Inference Engineering
AI & Machine Learning•Guide
Deep-dive technical guide to LLM inference optimization & runtime mechanics
