FFLAS-FFPACK - Finite Field Linear Algebra Subroutines / Package
C++ M4 C Makefile Shell
ClementPernet Merge pull request #168 from linbox-team/simd512
Let fflas-simd machinery support AVX512
Latest commit f4de66d Jul 20, 2018

README.md

FFLAS-FFPACK: Finite Field Linear Algebra Subroutines/Package

CI Inria: Build Status

Travis: Build Status

PURPOSE

The FFLAS-FFPACK library provides a set of basic routines for linear algebra over a finite field or the ring of integers with dense and sparse matrices.

It is inspired by the BLAS interface (Basic Linear Algebra Subprograms) and the LAPACK library for numerical linear algebra, and shares part of their design. Yet it differs in many aspects due to the specifities of computing over exact domains such as a finite fields and the field of rationals:

  • it is generic with respect to the finite field, so as to accomodate a large variety of field sizes and implementations;
  • consequently all routines use the C++ template genericity and the library is primarily meant to be used as a source code library, to be included and compiled in the user's software.
  • However, we also provide a compiled version instantiating most common routines over the most common finite fields.

LICENSE

FFLAS-FFPACK is distributed unded the terms of the GNU LGPL v2.1 or later (see LICENSE).

REQUIREMENTS:

  • a C++ compiler supporting C++11 standard. This means g++ v4.7 or greater, clang++ v3.4 or greater, icpc v16 or greater (earlier versions of clang and icpc might also work but have not been tested)
  • A BLAS library conforming to either the C or Fortran BLAS standard: OpenBLAS (recommended), or ATLAS. Make sure to use a single threaded version of the BLAS library.
  • Givaro version at least 4.0.1, providing the implementations of the coefficient fields/rings.

INSTALLATION

In brief: ./configure <options> && make && make install

The most commonly used option include:

  • --with-blas-libs=<libs> : to specify the arguments for the linker to find the BLAS
  • --enable-precompilation : to precompile the standard templates specializations (and gain some compilation time later on)

Type ./configure --help to list all options available. Note that givaro is automatically detected by pkg-config, so you no longer need to pass a --with-givaro=... option. You may need to set the PKG_CONFIG_PATH environment variable to <givaro-prefix>/lib/pkgconfig if you have installed it in a non standard directory.

For example on a x86_64 architecture:

  • Using OpenBLAS in Fedora:
  • install the package openblas-devel.x86_64,
  • run ./configure --with-blas-libs="-lopenblas"
  • Using OpenBLAS in Debian, Ubuntu, Mint, and all debian based distribution:
  • avoid using the distribution's package, as it is threaded by default. You need to compile openblas yourself on these systems,
  • run ./configure --with-blas-libs="-lopenblas"
  • Using ATLAS in Debian, Ubuntu, Mint:
  • install the package libatlas-dev,
  • run ./configure --with-blas-libs="-latlas -lcblas"
  • Using ATLAS in Fedora:
  • install the package atlas-devel.x86_64,
  • run ./configure --with-blas-libs="-L/usr/lib64/atlas -lsatlas".
  • Using Accelerate Framework on OS-X:
  • run ./configure --with-blas-libs="-framework Accelerate".
  • Using BLIS
  • Configure BLIS with, say, ./configure --enable-cblas auto.
  • run fflas/ffpack's ./configure --with-blas-libs="-lblis -lpthread".

Then, simply run make; make autotune; make install; make check Note that running the autotune target is optional but recommended as it will tune up the thresholds of various algorithms to your specific target host. make check is also optional but recommended as a sanity check.

see INSTALL for further details.

KNOWN BUGS

AVAILABILITY

from linbox-team/fflas-ffpack

AUTHORS

The FFLAS-FFPACK group (see AUTHORS file for a list of contributors).

Citing FFLAS-FFPACK

If your research depends on the FFLAS-FFPACK library, please consider citing the project as

@manual{fflas-ffpack,
title = {{FFLAS-FFPACK}: {F}inite {F}ield {L}inear {A}lgebra {S}ubroutines / {P}ackage},
author = {The FFLAS-FFPACK group},
edition = {v2.2.1},
year = {2016},
note = {\url{http://github.com/linbox-team/fflas-ffpack}}
}

Or you may also consider citing the related research article:

@article{DGP:2008,
author = {Jean-Guillaume Dumas and Pascal Giorgi and Cl{\'e}ment Pernet},
title = {Dense Linear Algebra over Word-Size Prime Fields: the FFLAS and FFPACK Packages},
journal = {ACM Trans. on Mathematical Software (TOMS)},
volume = {35},
number = {3},
year = {2008},
issn = {0098-3500},
pages = {1--42},
doi = {10.1145/1391989.1391992},
publisher = {ACM Press},
address = {New York, NY, USA}
}

Contact and discussion

For any bug report, feature or help request, please file an issue on github's issue tracker.

Please address any other request, suggestion and comment to the discussion group ffpack-devel.