cuda-python

CUDA Python is the home for accessing NVIDIA’s CUDA platform from Python. It consists of multiple components:

• cuda.core: Pythonic access to CUDA Runtime and other core functionality
• cuda.bindings: Low-level Python bindings to CUDA C APIs
• cuda.pathfinder: Utilities for locating CUDA components installed in the user's Python environment
• cuda.compute: A Python module for easy access to CCCL's highly efficient and customizable parallel algorithms, like sort, scan, reduce, transform, etc. that are callable on the host
• numba-cuda-mlir: An evolution of Numba CUDA that improves upon its technical foundation and performance to provide the future of CUDA Python JIT compilation. It currently supports developing CUDA SIMT kernels in Python, providing Python bindings for accelerated device libraries, and serving as a compiler for user-defined functions in accelerated libraries.
• numba.cuda: A Python DSL that exposes CUDA SIMT programming model and compiles a restricted subset of Python code into CUDA kernels and device functions
• cuda.tile: A new Python DSL that exposes CUDA Tile programming model and allows users to write NumPy-like code in CUDA kernels
• nvmath-python: Pythonic access to NVIDIA CPU & GPU Math Libraries, with host, device, and distributed APIs. It also provides low-level Python bindings to host C APIs (nvmath.bindings).
• nvshmem4py: Pythonic interface to the NVSHMEM library, enabling Python applications to leverage NVSHMEM's high-performance PGAS (Partitioned Global Address Space) programming model for GPU-accelerated computing
• Nsight Python: Python kernel profiling interface that automates performance analysis across multiple kernel configurations using NVIDIA Nsight Tools
• CUPTI Python: Python APIs for creation of profiling tools that target CUDA Python applications via the CUDA Profiling Tools Interface (CUPTI)
• Accelerated Computing Hub: Open-source learning materials related to GPU computing. You will find user guides, tutorials, and other works freely available for all learners interested in GPU computing.

CUDA Python is currently undergoing an overhaul to improve existing and introduce new components. All of the previously available functionality from the cuda-python package will continue to be available, please refer to the cuda.bindings documentation for installation guide and further detail.

cuda-python as a metapackage

cuda-python is being restructured to become a metapackage that contains a collection of subpackages. Each subpackage is versioned independently, allowing installation of each component as needed.

Subpackage: cuda.core

The cuda.core package offers idiomatic, Pythonic access to CUDA Runtime and other functionalities.

The goals are to

1. Provide idiomatic ("Pythonic") access to CUDA Driver, Runtime, and JIT compiler toolchain
2. Focus on developer productivity by ensuring end-to-end CUDA development can be performed quickly and entirely in Python
3. Avoid homegrown Python abstractions for CUDA for new Python GPU libraries starting from scratch
4. Ease developer burden of maintaining and catching up with latest CUDA features
5. Flatten the learning curve for current and future generations of CUDA developers

Subpackage: cuda.bindings

The cuda.bindings package is a standard set of low-level interfaces, providing full coverage of and access to the CUDA host APIs from Python.

The list of available interfaces is:

• CUDA Driver
• CUDA Runtime
• NVRTC
• nvJitLink
• NVVM
• nvFatbin
• cuFile
• NVML

License

CUDA Python is licensed under the Apache License 2.0. Third-party attributions for cuda.core are listed in cuda_core/NOTICE.

Each subproject is distributed as its own package and ships a copy of the same license alongside its sources, so that the license accompanies the built wheel. The root LICENSE governs the repository as a whole: