1001 Ways to Write CUDA Kernels in Python | NVIDIA GTC 2025 Session

Описание: You have to write a CUDA kernel. Did you know you can do it without leaving Python? We'll explore best practices for writing CUDA kernels using Python, empowering developers to harness the full potential of GPU acceleration.

Gain a clear understanding of the structure and functionality of CUDA kernels, learning how to effectively implement them within Python applications. We’ll cover essential concepts including memory management, thread organization, and optimization techniques that are critical for achieving high performance.

Speaker: Leo Fang, Python CUDA Tech Lead, NVIDIA

Watch more NVIDIA GTC sessions on demand: https://www.nvidia.com/en-us/on-deman...

CUDA Toolkit: https://developer.nvidia.com/cuda-too...

Topic: Development and Optimization - Programming Languages / Compilers
Level: General Interest
NVIDIA technology: CUDA,CUDA-X

Replay of NVIDIA GTC 2025 session S72449.