MPI support

Detailed Description

MPI support for nda::basic_array and nda::basic_array_view objects.

nda uses the TRIQS/mpi library to provide functions to broadcast, gather, reduce and scatter arrays and views over MPI processes.

The following example demonstrates some of these features:

#include <nda/mpi.hpp>
#include <nda/nda.hpp>
 
#include <iostream>
#include <mpi/mpi.hpp>
 
int main(int argc, char **argv) {
  // initialize MPI environment
  mpi::environment env(argc, argv);
  mpi::communicator comm;
 
  // create a 2x2 array on each process and fill it with its rank
  nda::array<int, 2> A(2, 2);
  A() = comm.rank();
 
  // reduce the array over all processes, which returns an mpi::lazy proxy object
  auto lazy_sum = mpi::reduce(A);
 
  // since it satisfies the nda::ArrayInitializer concept, we can use it to initialize an nda::array
  nda::array<int, 2> sum(lazy_sum);
 
  // print the result
  if (comm.rank() == 0) std::cout << sum << std::endl;
}

Running with 4 cores outputs:

[[6,6]
 [6,6]]

Classes
struct	mpi::lazy< mpi::tag::gather, A >
	Specialization of the mpi::lazy class for nda::Array types and the mpi::tag::gather tag. More...
struct	mpi::lazy< mpi::tag::reduce, A >
	Specialization of the mpi::lazy class for nda::Array types and the mpi::tag::reduce tag. More...
struct	mpi::lazy< mpi::tag::scatter, A >
	Specialization of the mpi::lazy class for nda::Array types and the mpi::tag::scatter tag. More...

Functions
template<typename A > requires (is_regular_or_view_v<A>)
void	nda::mpi_broadcast (A &a, mpi::communicator comm={}, int root=0)
	Implementation of an MPI broadcast for nda::basic_array or nda::basic_array_view types.
template<typename A > requires (is_regular_or_view_v<A>)
ArrayInitializer< std::remove_reference_t< A > > auto	nda::mpi_gather (A &&a, mpi::communicator comm={}, int root=0, bool all=false)
	Implementation of an MPI gather for nda::basic_array or nda::basic_array_view types.
template<typename A > requires (is_regular_or_view_v<A>)
ArrayInitializer< std::remove_reference_t< A > > auto	nda::mpi_reduce (A &&a, mpi::communicator comm={}, int root=0, bool all=false, MPI_Op op=MPI_SUM)
	Implementation of an MPI reduce for nda::basic_array or nda::basic_array_view types.
template<typename A > requires (is_regular_or_view_v<A>)
ArrayInitializer< std::remove_reference_t< A > > auto	nda::mpi_scatter (A &&a, mpi::communicator comm={}, int root=0, bool all=false)
	Implementation of an MPI scatter for nda::basic_array or nda::basic_array_view types.

Function Documentation

mpi_broadcast()

template<typename A >
requires (is_regular_or_view_v<A>)

void nda::mpi_broadcast	(	A &	a,
		mpi::communicator	comm = {},
		int	root = 0 )

#include <nda/mpi/broadcast.hpp>

Implementation of an MPI broadcast for nda::basic_array or nda::basic_array_view types.

For the root process, the array/view is broadcasted to all other processes. For non-root processes, the array/view is resized/checked to match the broadcasted dimensions and the data is written into the given array/view.

Throws an exception, if a given view does not have the correct shape.

// create an array on all processes
nda::array<int, 2> arr(3, 4);
 
// ...
// fill array on root process
// ...
 
// broadcast the array to all processes
mpi::broadcast(arr);

Template Parameters

A	nda::basic_array or nda::basic_array_view type.

Parameters

a	Array or view to be broadcasted from/into.
comm	mpi::communicator object.
root	Rank of the root process.

Definition at line 61 of file broadcast.hpp.

mpi_gather()

template<typename A >
requires (is_regular_or_view_v<A>)

ArrayInitializer< std::remove_reference_t< A > > auto nda::mpi_gather	(	A &&	a,
		mpi::communicator	comm = {},
		int	root = 0,
		bool	all = false )

#include <nda/mpi/gather.hpp>

Implementation of an MPI gather for nda::basic_array or nda::basic_array_view types.

Since the returned mpi::lazy object models an nda::ArrayInitializer, it can be used to initialize/assign to nda::basic_array and nda::basic_array_view objects:

// create an array on all processes
nda::array<int, 2> arr(3, 4);
 
// ...
// fill array on each process
// ...
 
// gather the array to the root process
nda::array<int, 2> res = mpi::gather(arr);

Here, the array res will have a shape of (3 * comm.size(), 4).

Template Parameters

A	nda::basic_array or nda::basic_array_view type.

Parameters

a	Array or view to be gathered.
comm	mpi::communicator object.
root	Rank of the root process.
all	Should all processes receive the result of the gather.

Returns

An mpi::lazy object modelling an nda::ArrayInitializer.

Definition at line 166 of file gather.hpp.

mpi_reduce()

template<typename A >
requires (is_regular_or_view_v<A>)

ArrayInitializer< std::remove_reference_t< A > > auto nda::mpi_reduce	(	A &&	a,
		mpi::communicator	comm = {},
		int	root = 0,
		bool	all = false,
		MPI_Op	op = MPI_SUM )

#include <nda/mpi/reduce.hpp>

Implementation of an MPI reduce for nda::basic_array or nda::basic_array_view types.

Since the returned mpi::lazy object models an nda::ArrayInitializer, it can be used to initialize/assign to nda::basic_array and nda::basic_array_view objects:

// create an array on all processes
nda::array<int, 2> arr(3, 4);
 
// ...
// fill array on each process
// ...
 
// reduce the array to the root process
nda::array<int, 2> res = mpi::reduce(arr);

Template Parameters

A	nda::basic_array or nda::basic_array_view type.

Parameters

a	Array or view to be gathered.
comm	mpi::communicator object.
root	Rank of the root process.
all	Should all processes receive the result of the gather.
op	MPI reduction operation.

Returns

An mpi::lazy object modelling an nda::ArrayInitializer.

Definition at line 165 of file reduce.hpp.

mpi_scatter()

template<typename A >
requires (is_regular_or_view_v<A>)

ArrayInitializer< std::remove_reference_t< A > > auto nda::mpi_scatter	(	A &&	a,
		mpi::communicator	comm = {},
		int	root = 0,
		bool	all = false )

#include <nda/mpi/scatter.hpp>

Implementation of an MPI scatter for nda::basic_array or nda::basic_array_view types.

Since the returned mpi::lazy object models an nda::ArrayInitializer, it can be used to initialize/assign to nda::basic_array and nda::basic_array_view objects:

// create an array on all processes
nda::array<int, 2> arr(10, 4);
 
// ...
// fill array on root process
// ...
 
// scatter the array to all processes
nda::array<int, 2> res = mpi::scatter(arr);

Here, the array res will have a shape of (10 / comm.size(), 4).

Template Parameters

A	nda::basic_array or nda::basic_array_view type.

Parameters

a	Array or view to be scattered.
comm	mpi::communicator object.
root	Rank of the root process.
all	Should all processes receive the result of the scatter (not used).

Returns

An mpi::lazy object modelling an nda::ArrayInitializer.

Definition at line 156 of file scatter.hpp.