GPU Programming in Pure Python

This talk from PyCon US provides a beginner-friendly introduction to GPU programming using Python. Learn how the CUDA model works and discover how to manage accelerator devices through the cuda.core interface. Follow along with example-driven demonstrations that cover launching work, managing memory, implementing parallel algorithms with cuda.parallel, writing custom kernels with cuda.cooperative, and integrating with accelerated libraries like cuDNN and cuBLAS. Explore practical parallel programming examples ranging from word counting to softmax implementation and a complete machine learning demonstration. In just 48 minutes, gain the knowledge needed to start accelerating your Python code with GPUs without the intimidation factor typically associated with GPU programming.