Glossary · Phase 7 · Software engineering
CUDA kernel
A function that runs on an NVIDIA GPU, once per thread; each thread works out which element is its own from its block and thread numbers.
Try it now
Spot the language from one or two telltale marks: that's often all you need to start reading.
Learn it properly
Reading C, C++, CUDA and Rust · Pointers and loops in C, references and templates in C++, one-thread-per-element CUDA kernels, and Rust's Option, Result and ownership. About 30 minutes.
Other terms from the same lesson
- Rust: A compiled systems language that frees memory safely without a garbage collector, through ownership and borrowing. uv, Ruff and Hugging Face's tokenizers are written in it.
Never trained an AI?
Teach a small one to write, right in your browser. It takes about 20 seconds and needs no account.