CUDA Documents
Official references, written tutorials, and sample code worth keeping close while you learn.
CUDA C++ Programming Guide
The official reference for the CUDA programming model, language extensions, and runtime.
Open resource →Official docsCUDA C++ Best Practices Guide
Practical guidance for memory, execution, and instruction-level optimization.
Open resource →Official docsCUDA Quick Start Guide
A fast path for installing CUDA and verifying your local setup.
Open resource →Hands-on codeCUDA Samples Repository
Official NVIDIA sample code for learning by example and testing APIs.
Open resource →NVIDIA blogAdvanced Strategies for High-Performance GPU Programming
NVIDIA Technical Blog post with an expert CUDA performance talk.
Open resource →How to use these references
Start with the CUDA Quick Start Guide to get a working toolkit, then read the first chapters of the CUDA C++ Programming Guide alongside our introduction and programming model chapters. The Programming Guide is the authoritative reference for kernel syntax, the thread hierarchy and the memory model; keep it open when a tutorial mentions an API you have not seen before.
Once your kernels work, the CUDA C++ Best Practices Guide explains how to make them fast: memory coalescing, occupancy and measuring with a profiler. It pairs with our performance optimization chapter. The CUDA Samples repository is the quickest way to see complete, compilable programs; you can paste most single-file samples into the online playground and run them on a real GPU without installing anything.