Skip to content
/ rocFFT Public
forked from ROCm/rocFFT

Next generation FFT implementation for ROCm

License

Notifications You must be signed in to change notification settings

amdkila/rocFFT

 
 

Repository files navigation

rocFFT

rocFFT is a software library for computing Fast Fourier Transforms (FFT) written in HIP. It is part of AMD's software ecosystem based on ROCm. In addition to AMD GPU devices, the library can also be compiled with the CUDA compiler using HIP tools for running on Nvidia GPU devices.

Installing pre-built packages

Download pre-built packages either from ROCm's package servers or by clicking the github releases tab and downloading the source, which could be more recent than the pre-build packages. Release notes are available for each release on the releases tab.

  • sudo apt update && sudo apt install rocfft

Building from source

rocFFT is compiled with hcc and uses cmake. To compile the library one calls, for example, the following commands:

mkdir build && cd build
cmake -DCMAKE_CXX_COMPILER=hcc .. 

A static library can be compiled by using the option -DBUILD_SHARED_LIBS=off

rocfft-rider is a client which will run general transforms and is useful for performance analysis. Compilation is enabled via the -DBUILD_CLIENTS_RIDER=on cmake option. rocfft-rider uses boost program options.

rocfft-test runs functionality tests and uses FFTW, Google test, and boost program options. Compilation is enabled by calling cmake with the -DBUILD_CLIENTS_TESTS=on option.

To install the clients depencencies on Ubuntu, run sudo apt install libgtest-dev libfftw3 libboost-program-options-dev

The file install.sh is a bash script is a wrapper for the cmake script, which also install dependencies on certain Linux distributions. The preferred method for compiling rocFFT is to call cmake directly.

Library and API Documentation

Please refer to the Library documentation for current documentation.

Examples

Examples may be found in the clients/samples subdirectory.

About

Next generation FFT implementation for ROCm

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • C++ 79.8%
  • Python 10.7%
  • CMake 4.4%
  • C 2.2%
  • Asymptote 1.3%
  • Shell 1.2%
  • Groovy 0.4%