Geek's blog about optimization, architectures and everything.
This is a list of presentations for my course about NVIDIA GPU architecture. The main aims of this course are
The first presentation explains key properties of the architecture. The second one describes the concepts of CUDA's programming, execution and memory models. The third presentation highlights the main optimization principle. The fourth presentation describes the memory architecture in details. The fifth gives concrete pieces of advice for optimizers about dealing with SIMT GPUs. The sixth presentation gives more concrete examples of optimization approaches. The last one describes how to identify performance limiters.