LLM Systems 05: Deep Learning Framework and Auto Differentiation
0. Recap
Tiling for efficient matrix computation
Use shared memory to reduce global memory access
Roofline Model & Computational Intensity
Coalesced memory access
Understanding the ch