parVectorMap< S > Class Template Reference

A class which determines the way to distribute a vector across MPI procs. More...

#include <parVectorMap.hpp>

Public Member Functions
	parVectorMap ()
	A constructor of `parVectorMap`.
	parVectorMap (MPI_Comm ncomm, S lbound, S ubound)
	A constructor of `parVectorMap`. More...
	~parVectorMap ()
	A destructor of `parVectorMap`.
bool	operator== (const parVectorMap &map1)
	Compare if this map is identical to another one `map1`.
bool	operator!= (const parVectorMap &map1)
	Compare if this map is different with another one `map1`.
S	Loc2Glob (S local_index)
	Convert a index of local vector on each MPI proc into its index in the global distributed vector. More...
S	Glob2Loc (S global_index)
	Convert a index of global vector into its index in the local vector on each MPI proc. More...
MPI_Comm	GetCurrentComm ()
	Return parVectorMap<S>::comm.
int	GetRank ()
	Return parVectorMap<S>::rank.
S	GetLowerBound ()
	Return parVectorMap<S>::lower_bound.
S	GetUpperBound ()
	Return parVectorMap<S>::upper_bound.
S	GetLocalSize ()
	Return parVectorMap<S>::local_size.
S	GetGlobalSize ()
	Return parVectorMap<S>::global_size.
std::vector< S >	GetLBoundMap ()
	Return parVectorMap<S>::lprocbound_map.
std::vector< S >	GetUBoundMap ()
	Return parVectorMap<S>::uprocbound_map.

Private Attributes
MPI_Comm	comm
	The working MPI Communicator.
int	nproc
	number of MPI procs within the working MPI communicator parVectorMap::comm
int	rank
	rank of each MPI procs within the working MPI communicator parVectorMap::comm
S	lower_bound
	the smallest index of a distributed vector on each MPI proc
S	upper_bound
	`upper_bound-1 = ` the largest index of a distributed vector on each MPI proc
S	local_size
	The number of elements of vector stored on each MPI proc.
S	global_size
	Global size of this distributed vector.
std::vector< S >	lprocbound_map
	A `std::vector` which stores the parVectorMap::lower_bound of all MPI procs together.
std::vector< S >	uprocbound_map
	A `std::vector` which stores the parVectorMap::upper_bound of all MPI procs together.

Detailed Description

template<typename S>
class parVectorMap< S >

A class which determines the way to distribute a vector across MPI procs.

• This class is to create a mapping from a fixed-size vector to multiple MPI procs in 1D grid.
• This class can also be used to create more distributed vectors and sparse matrices following the same way.
• For each MPI proc, a piece of vector with indexing `[lower_bound, upper_bound)` is allocated.

Template Parameters

S	type of integer to describes the dimension of vector to be generated.

Constructor & Destructor Documentation

parVectorMap()

template<typename S >

parVectorMap< S >::parVectorMap	(	MPI_Comm	ncomm,
		S	lbound,
		S	ubound
	)

A constructor of `parVectorMap`.

Parameters

[in]	ncomm	the working MPI Communicator
[in]	lbound	the smallest index of a distributed vector on each MPI proc
[in]	ubound	`ubound-1 = ` the largest index of a distributed vector on each MPI proc

Member Function Documentation

Glob2Loc()

template<typename S >

S parVectorMap< S >::Glob2Loc

(

S

global_index

)

Convert a index of global vector into its index in the local vector on each MPI proc.

Parameters

[in]

global_index

the index global vector to be converted

• Attention, this function is in distributed manner, each MPI proc can only convert the global index of vector in the range `[lower_bound, upper_bound)`.

Loc2Glob()

template<typename S >

S parVectorMap< S >::Loc2Glob

(

S

local_index

)

Convert a index of local vector on each MPI proc into its index in the global distributed vector.

Parameters

[in]

local_index

the index local vector to be converted

• Attention, this function is in distributed manner, each MPI proc can only convert the local index of vector stored on itself.

---

The documentation for this class was generated from the following file:

• include/parVector/parVectorMap.hpp