Skip to content

STEllAR-GROUP/libflatarray

 
 

Repository files navigation

ABOUT
=====

LibFlatArray acts as a highly efficient multi-dimensional array of
arbitrary objects (array of structs, AoS), but really uses a struct of
arrays (SoA) memory layout. It's great for writing vectorized code and
its lightning-fast iterators give you access to neighboring elements
with zero address generation overhead.

Use cases include:
- computer simulations (e.g. stencil codes such as Lattice Boltzmann Methods)
- image processing (e.g. Gaussian filters)
- numerical methods (e.g. multiplication of complex matrices)

The library is written in C++ and uses templates to shift the burden
of address computation from runtime to compile time. It shares some
infrastructure with its parent project LibGeoDecomp.

Further information:
  - homepage:          http://www.libgeodecomp.org/libflatarray.html
  - mailing list:      http://www.libgeodecomp.org/mailing_lists.html
  - source repository: https://bitbucket.org/gentryx/libflatarray
  - contributors:      see file "AUTHORS"

DEPENDENCIES
============

- C++ compiler (min. C++98, tested with GCC's g++, Clang's clang++,
  and Intel's icpc)

- CMake (min. 2.8.10)

- build tool supported by CMake (e.g. make, ninja)

BUILDING
========

For compiling LibFlatArray you'll need CMake (http://www.cmake.org)
installed. We recommend an out-of-source build:

  BUILD_DIR=build/`uname -ms | sed s/\ /-/g`
  mkdir -p $BUILD_DIR
  cd $BUILD_DIR
  cmake ../../
  make

About

Multi-dimensional C++ arrays which store objects in a Struct-of-Arrays (SoA) memory layout for efficient vectorization and zero address generation overhead.

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • C++ 96.2%
  • Cuda 2.4%
  • CMake 1.4%