Skip to content
/ nda Public
forked from TRIQS/nda

C++ library for multi-dimensional arrays

License

Notifications You must be signed in to change notification settings

jasonkaye/nda

 
 

Repository files navigation

build

nda - A multi-dimensional array library in C++

nda is a C++ library providing an efficient and flexible multi-dimensional array class. It is an essential building-block of the TRIQS project. Some features include

  • coded in C++17/20 using concepts
  • expressions are implemented lazily for maximum performance
  • flexible and lightweight view-types
  • matrix and vector class with BLAS / LAPACK backend
  • easily store and retrieve arrays to and from hdf5 files using h5
  • common mpi functionality using mpi

Install instructions are provided here.

A prelimenary reference documentation based on Doxygen is provided at triqs.github.io/nda.

Simple Example

#include <nda/nda.hpp>
#include <nda/h5.hpp>

using namespace nda;

int main() {

  // Create array of shape (4,4,4)
  array<long, 3> A(4, 4, 4);

  // Create an array given its data
  array<long, 2> B{{1, 2}, {3, 4}, {5, 6}};

  // Assign
  A() = 0;
  A(0, 1, 2) = 40;

  // Access
  int a = A(0, 1, 2) + B(0, 1);

  // Slicing to a view of shape (3, 2)
  auto V = A(range(0, 3), range(0, 2), 0);

  // Lazy Arithmetic operations
  auto C = V + 2 * B;

  // Algorithms
  min_element(V);
  max_element(V);
  sum(V);

  // write to file
  {
    h5::file file("dat.h5", 'w');
    h5_write(file, "A", A);
  }

  // read from file
  array<long, 3> D;
  {
    h5::file file("dat.h5", 'r');
    h5_read(file, "A", D);
  }

}

For further examples we refer the users to our tests. We outline some important design principles here.

About

C++ library for multi-dimensional arrays

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • C++ 93.6%
  • CMake 3.1%
  • C 1.8%
  • Python 1.3%
  • Other 0.2%