| 1 |
Source: mpi4py-fft |
| 2 |
Section: python |
| 3 |
Priority: optional |
| 4 |
Maintainer: Debian Science Maintainers <[email protected]> |
| 5 |
Uploaders: Drew Parsons <[email protected]> |
| 6 |
Build-Depends: debhelper-compat (= 13), |
| 7 |
cython3, |
| 8 |
dh-python, |
| 9 |
libfftw3-dev, |
| 10 |
mpi-default-dev, |
| 11 |
python3-all-dev, |
| 12 |
python3-setuptools, |
| 13 |
python3-mpi4py, |
| 14 |
python3-numpy, |
| 15 |
python3-sphinx, |
| 16 |
Build-Depends-Indep: |
| 17 |
libjs-mathjax <!nodoc>, |
| 18 |
python3-sphinx-rtd-theme <!nodoc>, |
| 19 |
texinfo <!nodoc> |
| 20 |
Standards-Version: 4.6.2 |
| 21 |
Homepage: https://github.com/mpi4py/mpi4py-fft |
| 22 |
Vcs-Browser: https://salsa.debian.org/science-team/mpi4py-fft |
| 23 |
Vcs-Git: https://salsa.debian.org/science-team/mpi4py-fft.git |
| 24 |
|
| 25 |
Package: python3-mpi4py-fft |
| 26 |
Architecture: any |
| 27 |
Depends: ${python3:Depends}, ${shlibs:Depends}, ${misc:Depends} |
| 28 |
Suggests: python-mpi4py-fft-doc |
| 29 |
Description: a Python package for computing Fast Fourier Transforms (FFTs) with MPI |
| 30 |
mpi4py-fft is a Python package for computing Fast Fourier Transforms |
| 31 |
(FFTs). Large arrays are distributed and communications are handled |
| 32 |
under the hood by MPI for Python (mpi4py). To distribute large arrays |
| 33 |
we are using a new and completely generic algorithm that allows for |
| 34 |
any index set of a multidimensional array to be distributed. We can |
| 35 |
distribute just one index (a slab decomposition), two index sets |
| 36 |
(pencil decomposition) or even more for higher-dimensional arrays. |
| 37 |
. |
| 38 |
In mpi4py-fft there is also included a Python interface to the FFTW |
| 39 |
library. This interface can be used without MPI, much like pyfftw, |
| 40 |
and even for real-to-real transforms, like discrete cosine or sine |
| 41 |
transforms. |
| 42 |
. |
| 43 |
The package provides a Python interface to FFTW, but with MPI |
| 44 |
parallelisation. This enables strong scaling tested to 16000 cores, |
| 45 |
or weak scaling tested to 2000 cores. The algorithm is documented at |
| 46 |
https://arxiv.org/abs/1804.09536 |
| 47 |
. |
| 48 |
This package installs the library for Python 3. |
| 49 |
|
| 50 |
Package: python3-mpi4py-fft-doc |
| 51 |
Architecture: all |
| 52 |
Multi-Arch: foreign |
| 53 |
Section: doc |
| 54 |
Depends: libjs-mathjax, ${sphinxdoc:Depends}, ${misc:Depends} |
| 55 |
Description: a Python package for computing Fast Fourier Transforms (FFTs) with MPI (docs) |
| 56 |
mpi4py-fft is a Python package for computing Fast Fourier Transforms |
| 57 |
(FFTs). Large arrays are distributed and communications are handled |
| 58 |
under the hood by MPI for Python (mpi4py). To distribute large arrays |
| 59 |
we are using a new and completely generic algorithm that allows for |
| 60 |
any index set of a multidimensional array to be distributed. We can |
| 61 |
distribute just one index (a slab decomposition), two index sets |
| 62 |
(pencil decomposition) or even more for higher-dimensional arrays. |
| 63 |
. |
| 64 |
In mpi4py-fft there is also included a Python interface to the FFTW |
| 65 |
library. This interface can be used without MPI, much like pyfftw, |
| 66 |
and even for real-to-real transforms, like discrete cosine or sine |
| 67 |
transforms. |
| 68 |
. |
| 69 |
The package provides a Python interface to FFTW, but with MPI |
| 70 |
parallelisation. This enables strong scaling tested to 16000 cores, |
| 71 |
or weak scaling tested to 2000 cores. The algorithm is documented at |
| 72 |
https://arxiv.org/abs/1804.09536 |
| 73 |
. |
| 74 |
This is the common documentation package. |