Technical requirements

To be able to run the provided example codes in this chapter, you will need a system equipped with an NVIDIA GPU. We use Pixi as a dependency management tool that installs all required packages, including the Numba-CUDA library. To set up the environment, simply run pixi install in the root directory of the book's GitHub repository. This will ensure that all dependencies are installed. Once the installation is complete, you can activate the environment using pixi shell.

To use JupyterLab, run the following command:

pixi run jupyter-lab

You will also need to install NVIDIA Nsight Systems separately to enable profiling and visualization of the CUDA stream timeline. You can always get the latest version from NVIDIA's official Developer site (https://developer.nvidia.com/tools-downloads) available for different platforms such as Linux, Windows, and macOS. To obtain NVIDIA Nsight Compute metrics, we rely only on the CLI outputs. However, Nsight Compute can...