PTX

Instead of [CUDA](/en/CUDA), [DeepSeek](/en/DeepSeek) uses an intermediate language called PTX ([Parallel Thread Execution](/en/Parallel%20Thread%20Execution)) provided by NVIDIA. PTX is a [low-level](/en/low-level) [intermediate representation](/en/intermediate%20representation) generated from CUDA C/C++ code, which allows optimization in a form similar to GPU [machine language](/en/machine%20language), thus enabling finer hardware control.

---

This page is auto-translated from /nishio/PTX using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I'm very happy to spread my thought to non-Japanese readers.