m_tensorprod_shared::sharedmemorywindow Type Reference

Type for managing a shared memory array using MPI shared memory windows. More...

Public Member Functions
generic	init (this, nr_elements, domain)
	Initialize a array in shared memory using a MPI shared memory window.
generic	init (this, nr_elements, comm, my_rank)
	Initialize a array in shared memory using a MPI shared memory window.
procedure	destroy (this)
	Destroy a shared memory array and free the associated MPI window.
procedure	fence (this)
	Synchronize access to the shared memory window.
procedure	leader (this)
	Check if this process is the leader in the shared memory communicator.
procedure	cat (this, other)
	Concatenate two shared memory windows (not implemented)

Public Attributes
real(wp), dimension(:), pointer, contiguous	window => null()
	Pointer to the shared memory array.
integer	win_shmem_id
	MPI window identifier for the shared memory window.
integer	comm
	MPI communicator for the shared memory domain.
integer	my_rank
	Rank of this process in the shared memory communicator.
integer	nr_ranks
	Number of ranks sharing this memory (size of comm)

Detailed Description

Type for managing a shared memory array using MPI shared memory windows.

Member Function/Subroutine Documentation

cat()

procedure m_tensorprod_shared::sharedmemorywindow::cat	(	class(sharedmemorywindow), intent(inout)	this,
		class(sharedmemorywindow), intent(in)	other )

Concatenate two shared memory windows (not implemented)

Parameters

[in,out]	this	SharedMemoryWindow object to be concatenated to (output is [this, other])
[in]	other	SharedMemoryWindow object to concatenate from

destroy()

procedure m_tensorprod_shared::sharedmemorywindow::destroy

(

class(sharedmemorywindow), intent(inout)

this

)

Destroy a shared memory array and free the associated MPI window.

Parameters

[in,out]

this

SharedMemoryWindow object

fence()

procedure m_tensorprod_shared::sharedmemorywindow::fence

(

class(sharedmemorywindow), intent(inout)

this

)

Synchronize access to the shared memory window.

Parameters

[in,out]

this

SharedMemoryWindow object

init() [1/2]

generic m_tensorprod_shared::sharedmemorywindow::init	(	class(sharedmemorywindow), intent(inout)	this,
		integer, intent(in)	nr_elements,
		integer, intent(in)	comm,
		integer, intent(in)	my_rank )

Initialize a array in shared memory using a MPI shared memory window.

Parameters

[out]	this	SharedMemoryWindow object
[in]	nr_elements	Number of elements in the shared memory array
[in]	comm	MPI communicator for the shared memory domain (each rank on this communicator must be on the same node)
[in]	my_rank	Rank of this process in the shared memory communicator

init() [2/2]

generic m_tensorprod_shared::sharedmemorywindow::init	(	class(sharedmemorywindow), intent(inout)	this,
		integer, intent(in)	nr_elements,
		type(tensorproddomain), intent(in)	domain )

Initialize a array in shared memory using a MPI shared memory window.

Parameters

[out]	this	SharedMemoryWindow object
[in]	nr_elements	Number of elements in the shared memory array
[in]	domain	TensorProdDomain object defining the shared memory communicator

Note

The memory is shared among all processes in the domaincomm_shmem MPI communicator

leader()

procedure m_tensorprod_shared::sharedmemorywindow::leader

(

class(sharedmemorywindow), intent(in)

this

)

Check if this process is the leader in the shared memory communicator.

Parameters

[in]

this

SharedMemoryWindow object

Returns

.true. if this process is the leader (rank 0), .false. otherwise

---

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

• /builds/rwr/bspline_feec/src/tensorprod/m_tensorprod_shared.f90