HIP Python

2023-06-23

1 min read

Applies to Linux

Advanced Micro Devices, Inc.

Welcome to the documentation pages for HIP Python!

HIP Python provides low-level Cython and Python® bindings for the HIP runtime, HIPRTC, multiple math libraries and the communication library RCCL, and further a CUDA® Python Interoperability layer that aims to simplify the porting of CUDA Python Python and Cython programs.

Note

This documentation has been generated based on ROCm™ version 5.7.1.

Spotlight

How to Install HIP Python

Learn how to identify the correct HIP Python packages to install for your ROCm™ installation, where to retrieve the packages and what options you have to install them.

• Installation

How to Use HIP Python

Learn how to use HIP Python’s interfaces in your Python or Cython program. We present a large number of examples that cover HIP and HIPRTC as well as multiple math libraries (HIPBLAS, HIPRAND, HIPFFT) and the communication library RCCL. Learn how to launch your own kernels and how the different libraries interact.

• Basic Usage and Examples

How to Port CUDA Python Applications

Learn how you can use HIP Python’s CUDA Python interoperability layer to port or even directly run CUDA Python applications on AMD GPUs. The chapter covers Python and Cython programs.

• CUDA® Python Interoperability

HIP Python’s Adapter Types

Learn about the datatypes that HIP Python uses to translate between C and Python and that are designed to ease interoperability with other packages such as NumPy and Numba.

• HIP Python’s Adapter Types

HIP Python’s Python API

The full list of HIP Python Python variables, classes and functions.

• hip.hip

• hip.hiprtc

• hip.hipblas

• hip.rccl

• hip.hiprand

• hip.hipfft

• hip.hipsparse

• hip.roctx

• hip.hipsolver

• hip._hip_helpers

• hip._util.types

The CUDA Python Interoperability Layer’s Python API

The full list of the CUDA Python interoperability layer’s Python variables, classes and functions.

• cuda.cuda

• cuda.cudart

• cuda.nvrtc