FFTW is a comprehensive collection of fast C routines for computing the Discrete Fourier Transform (DFT) and various special cases thereof, copyrighted by MIT and distributed under the GNU General Public License. Refer fftw.org for details.
FFTW can compute transforms of real and complex-values arrays of arbitrary size and dimension. AMD FFTW includes selective kernels and routines optimized for the AMD processors.
Highlights of AMD FFTW 3.1
- Feature ‘AMD application optimization layer’ that uplifts the performance of HPC and scientific applications
- Feature ‘Fast MPI transpose algorithm’ to speed up the distributed MPI FFT computations
- Feature ‘Top N planner’ that minimizes single-threaded run-to-run variations
- Support for building AMD FFTW library on Windows
- GCC compilation support for AMD processors based on the AMD “Zen3” core architecture
Highlights of AMD FFTW 3.0
- New fast planner that improves the time for various planning modes in general and OPATIENT mode in particular. It can be enabled through configure option “–enable-amd-fast-planner”
- Support for configure option “AMD_ARCH” to help cross compilation, taking various options (auto, znver1, znver2) depending on AMD processor generation
- Quad precision support is now included for AOCC clang compiler from version 10 onwards
- Improved handling of –enable-debug and “CC” options by ‘configure’ when –enable-amd-opt is enabled
The packages containing AMD FFTW binaries, examples and documentation are available in the Downloads section below.
Source code for AMD FFTW will be available shortly on GitHub https://github.com/amd/amd-fftw.
Refer here for prior versions of AMD FFTW documentation and downloads.