message_i_passing.f90 File Reference

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Functions
subroutine	mp_shift_im (msg, group, displ_in)
	Shift around the data in msg.
subroutine	mp_shift_i (msg, group, displ_in)
	Shift around the data in msg.
subroutine	mp_alltoall_i11v (sb, scount, sdispl, rb, rcount, rdispl, group)
	All-to-all data exchange, rank-1 data of different sizes.
subroutine	mp_alltoall_i22v (sb, scount, sdispl, rb, rcount, rdispl, group)
	All-to-all data exchange, rank-2 data of different sizes.
subroutine	mp_alltoall_i (sb, rb, count, group)
	All-to-all data exchange, rank 1 arrays, equal sizes.
subroutine	mp_alltoall_i22 (sb, rb, count, group)
	All-to-all data exchange, rank-2 arrays, equal sizes.
subroutine	mp_alltoall_i33 (sb, rb, count, group)
	All-to-all data exchange, rank-3 data with equal sizes.
subroutine	mp_alltoall_i44 (sb, rb, count, group)
	All-to-all data exchange, rank 4 data, equal sizes.
subroutine	mp_alltoall_i45 (sb, rb, count, group)
	All-to-all data exchange, rank-4 data to rank-5 data.
subroutine	mp_alltoall_i34 (sb, rb, count, group)
	All-to-all data exchange, rank-3 data to rank-4 data.
subroutine	mp_alltoall_i54 (sb, rb, count, group)
	All-to-all data exchange, rank-5 data to rank-4 data.
subroutine	mp_send_i (msg, dest, tag, gid)
	Send one datum to another process.
subroutine	mp_send_iv (msg, dest, tag, gid)
	Send rank-1 data to another process.
subroutine	mp_recv_i (msg, source, tag, gid)
	Receive one datum from another process.
subroutine	mp_recv_iv (msg, source, tag, gid)
	Receive rank-1 data from another process.
subroutine	mp_bcast_i (msg, source, gid)
	Broadcasts a datum to all processes.
subroutine	mp_bcast_iv (msg, source, gid)
	Broadcasts rank-1 data to all processes.
subroutine	mp_bcast_im (msg, source, gid)
	Broadcasts rank-2 data to all processes.
subroutine	mp_bcast_i3 (msg, source, gid)
	Broadcasts rank-3 data to all processes.
subroutine	mp_sum_i (msg, gid)
	Sums a datum from all processes with result left on all processes.
subroutine	mp_sum_iv (msg, gid)
	Element-wise sum of a rank-1 array on all processes.
subroutine	mp_sum_im (msg, gid)
	Element-wise sum of a rank-2 array on all processes.
subroutine	mp_sum_im3 (msg, gid)
	Element-wise sum of a rank-3 array on all processes.
subroutine	mp_sum_im4 (msg, gid)
	Element-wise sum of a rank-4 array on all processes.
subroutine	mp_sum_im5 (msg, gid)
	Element-wise sum of a rank-5 array on all processes.
subroutine	mp_sum_im6 (msg, gid)
	Element-wise sum of a rank-6 array on all processes.
subroutine	mp_sum_root_iv (msg, root, gid)
	Element-wise sum of data from all processes with result left only on one.
subroutine	mp_sum_root_im (msg, root, gid)
	Element-wise sum of data from all processes with result left only on one.
subroutine	mp_max_i (msg, gid)
	Finds the maximum of a datum with the result left on all processes.
subroutine	mp_max_iv (msg, gid)
	Finds the element-wise maximum of a vector with the result left on all processes.
subroutine	mp_min_i (msg, gid)
	Finds the minimum of a datum with the result left on all processes.
subroutine	mp_min_iv (msg, gid)
	Finds the element-wise minimum of vector with the result left on all processes.
subroutine	mp_scatter_iv (msg_scatter, msg, root, gid)
	Scatters data from one processes to all others.
subroutine	mp_gather_i (msg, msg_gather, root, gid)
	Gathers a datum from all processes to one.
subroutine	mp_gather_iv (msg, msg_gather, root, gid)
	Gathers data from all processes to one.
subroutine	mp_gather_im (msg, msg_gather, root, gid)
	Gathers data from all processes to one.
subroutine	mp_gatherv_iv (sendbuf, recvbuf, recvcounts, displs, root, comm)
	Gathers data from all processes to one.
subroutine	mp_allgather_i (msgout, msgin, gid)
	Gathers a datum from all processes and all processes receive the same data.
subroutine	mp_allgather_i12 (msgout, msgin, gid)
	Gathers vector data from all processes and all processes receive the same data.
subroutine	mp_allgather_i23 (msgout, msgin, gid)
	Gathers matrix data from all processes and all processes receive the same data.
subroutine	mp_allgather_i34 (msgout, msgin, gid)
	Gathers rank-3 data from all processes and all processes receive the same data.
subroutine	mp_allgatherv_iv (msgout, msgin, rcount, rdispl, gid)
	Gathers vector data from all processes and all processes receive the same data.
subroutine	mp_sum_scatter_iv (msgout, msgin, rcount, gid)
	Sums a vector and partitions the result among processes.
subroutine	mp_sendrecv_iv (msgin, dest, msgout, source, comm)
	Sends and receives vector data.
subroutine	mp_sendrecv_im2 (msgin, dest, msgout, source, comm)
	Sends and receives matrix data.
subroutine	mp_sendrecv_im3 (msgin, dest, msgout, source, comm)
	Sends and receives rank-3 data.
subroutine	mp_isendrecv_iv (msgin, dest, msgout, source, comm, send_request, recv_request, tag)
	Non-blocking send and receieve of a vector.
subroutine	mp_isendrecv_im2 (msgin, dest, msgout, source, comm, send_request, recv_request, tag)
	Non-blocking send and receieve of a matrix.
subroutine	mp_isend_iv (msgin, dest, comm, request, tag)
	Non-blocking send of vector data.
subroutine	mp_isend_im2 (msgin, dest, comm, request, tag)
	Non-blocking send of matrix data.
subroutine	mp_isend_im3 (msgin, dest, comm, request, tag)
	Non-blocking send of rank-3 data.
subroutine	mp_irecv_iv (msgout, source, comm, request, tag)
	Non-blocking receive of vector data.
subroutine	mp_irecv_im2 (msgout, source, comm, request, tag)
	Non-blocking receive of matrix data.
subroutine	mp_irecv_im3 (msgout, source, comm, request, tag)
	Non-blocking send of rank-3 data.
subroutine	mp_win_create_i (window, range, len, gid)
	Creates an MPI RMA window.
subroutine	mp_rma_get_i (window, remote, offset, len, storage)
	Fetch access using RMA.
subroutine	mp_rma_put_i (window, storage, len, remote, offset)
	Store access using RMA.
subroutine	mp_allocate_i (DATA, len, stat)
	Allocates special parallel memory.
subroutine	mp_deallocate_i (DATA, stat)
	Deallocates special parallel memory.

---

Function Documentation

subroutine mp_allgather_i	(	INTEGER(KIND=int_4),intent(in)	msgout,
		INTEGER(KIND=int_4),dimension( : ),intent(out)	msgin,
		INTEGER,intent(in)	gid
	)

Gathers a datum from all processes and all processes receive the same data.

Data size

All processes send equal-sized data

MPI mapping

mpi_allgather

Parameters:

[in]	msgout	Datum to send
[out]	msgin	Received data
[in]	gid	Message passing environment identifier

Definition at line 1544 of file message_i_passing.f90.

References timings::timeset().

Here is the call graph for this function:

subroutine mp_allgather_i12	(	INTEGER(KIND=int_4),dimension(:),intent(in)	msgout,
		INTEGER(KIND=int_4),dimension(:, :),intent(out)	msgin,
		INTEGER,intent(in)	gid
	)

Gathers vector data from all processes and all processes receive the same data.

Data size

All processes send equal-sized data

Ranks

The last rank counts the processes

MPI mapping

mpi_allgather

Parameters:

[in]	msgout	Rank-1 data to send
[out]	msgin	Received data
[in]	gid	Message passing environment identifier

Definition at line 1587 of file message_i_passing.f90.

References timings::timeset().

Here is the call graph for this function:

subroutine mp_allgather_i23	(	INTEGER(KIND=int_4),dimension(:,:),intent(in)	msgout,
		INTEGER(KIND=int_4),dimension(:, :, :),intent(out)	msgin,
		INTEGER,intent(in)	gid
	)

Gathers matrix data from all processes and all processes receive the same data.

Parameters:

[in]

msgout

Rank-2 data to send

See also:

mp_allgather_i12

Definition at line 1623 of file message_i_passing.f90.

References timings::timeset().

Here is the call graph for this function:

subroutine mp_allgather_i34	(	INTEGER(KIND=int_4),dimension(:,:, :),intent(in)	msgout,
		INTEGER(KIND=int_4),dimension(:, :, :, :),intent(out)	msgin,
		INTEGER,intent(in)	gid
	)

Gathers rank-3 data from all processes and all processes receive the same data.

Parameters:

[in]

msgout

Rank-3 data to send

See also:

mp_allgather_i12

Definition at line 1659 of file message_i_passing.f90.

References timings::timeset().

Here is the call graph for this function:

subroutine mp_allgatherv_iv	(	INTEGER(KIND=int_4),dimension( : ),intent(in)	msgout,
		INTEGER(KIND=int_4),dimension( : ),intent(out)	msgin,
		INTEGER,dimension( : ),intent(in)	rcount,
		INTEGER,dimension( : ),intent(in)	rdispl,
		INTEGER,intent(in)	gid
	)

Gathers vector data from all processes and all processes receive the same data.

Data size

Processes can send different-sized data

Ranks

The last rank counts the processes

Offsets

Offsets are from 0

MPI mapping

mpi_allgather

Parameters:

[in]	msgout	Rank-1 data to send
[out]	msgin	Received data
[in]	rcount	Size of sent data for every process
[in]	rdispl	Offset of sent data for every process
[in]	gid	Message passing environment identifier

Definition at line 1706 of file message_i_passing.f90.

References timings::timeset().

Here is the call graph for this function:

subroutine mp_allocate_i	(	INTEGER(KIND=int_4),dimension(:),pointer	DATA,
		INTEGER,intent(in)	len,
		INTEGER,intent(out),optional	stat
	)

Allocates special parallel memory.

Author:

UB

Parameters:

[in]	data	pointer to integer array to allocate
[in]	len	number of integers to allocate
[out]	stat	(optional) allocation status result

Definition at line 2560 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_alltoall_i	(	INTEGER(KIND=int_4),dimension(:),intent(in)	sb,
		INTEGER(KIND=int_4),dimension(:),intent(out)	rb,
		INTEGER,intent(in)	count,
		INTEGER,intent(in)	group
	)

All-to-all data exchange, rank 1 arrays, equal sizes.

Index meaning

The first two indices specify the data while the last index counts

the processes

Sizes of ranks

All processes have the same data size.

MPI mapping

mpi_alltoall

Parameters:

[in]	sb	array with data to send
[out]	rb	array into which data is received
[in]	count	number of elements to send/receive (product of the extents of the first two dimensions)
[in]	group	Message passing environment identifier

Definition at line 234 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_alltoall_i11v	(	INTEGER(KIND=int_4),dimension(:),intent(in)	sb,
		INTEGER,dimension(:),intent(in)	scount,
		INTEGER,dimension(:),intent(in)	sdispl,
		INTEGER(KIND=int_4),dimension(:),intent(inout)	rb,
		INTEGER,dimension(:),intent(in)	rcount,
		INTEGER,dimension(:),intent(in)	rdispl,
		INTEGER,intent(in)	group
	)

All-to-all data exchange, rank-1 data of different sizes.

MPI mapping

mpi_alltoallv

Array sizes

The scount, rcount, and the sdispl and rdispl arrays have a size equal to the number of processes.

Offsets

Values in sdispl and rdispl start with 0.

Parameters:

[in]	sb	Data to send
[in]	scount	Data counts for data sent to other processes
[in]	sdispl	Respective data offsets for data sent to process
[in,out]	rb	Buffer into which to receive data
[in]	rcount	Data counts for data received from other processes
[in]	rdispl	Respective data offsets for data received from other processes
[in]	group	Message passing environment identifier

Definition at line 139 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_alltoall_i22	(	INTEGER(KIND=int_4),dimension(:, :),intent(in)	sb,
		INTEGER(KIND=int_4),dimension(:, :),intent(out)	rb,
		INTEGER,intent(in)	count,
		INTEGER,intent(in)	group
	)

All-to-all data exchange, rank-2 arrays, equal sizes.

See also:

mp_alltoall_i

Definition at line 273 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_alltoall_i22v	(	INTEGER(KIND=int_4),dimension(:, :),intent(in)	sb,
		INTEGER,dimension(:),intent(in)	scount,
		INTEGER,dimension(:),intent(in)	sdispl,
		INTEGER(KIND=int_4),dimension(:, :),intent(inout)	rb,
		INTEGER,dimension(:),intent(in)	rcount,
		INTEGER,dimension(:),intent(in)	rdispl,
		INTEGER,intent(in)	group
	)

All-to-all data exchange, rank-2 data of different sizes.

MPI mapping

mpi_alltoallv

See also:

mp_alltoall_i11v

Definition at line 182 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_alltoall_i33	(	INTEGER(KIND=int_4),dimension(:, :, :),intent(in)	sb,
		INTEGER(KIND=int_4),dimension(:, :, :),intent(out)	rb,
		INTEGER,intent(in)	count,
		INTEGER,intent(in)	group
	)

All-to-all data exchange, rank-3 data with equal sizes.

See also:

mp_alltoall_i

Definition at line 312 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_alltoall_i34	(	INTEGER(KIND=int_4),dimension(:, :, :),intent(in)	sb,
		INTEGER(KIND=int_4),dimension(:, :, :, :),intent(out)	rb,
		INTEGER,intent(in)	count,
		INTEGER,intent(in)	group
	)

All-to-all data exchange, rank-3 data to rank-4 data.

Note:

User must ensure size consistency.

See also:

mp_alltoall_i

Definition at line 434 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_alltoall_i44	(	INTEGER(KIND=int_4),dimension(:, :, :, :),intent(in)	sb,
		INTEGER(KIND=int_4),dimension(:, :, :, :),intent(out)	rb,
		INTEGER,intent(in)	count,
		INTEGER,intent(in)	group
	)

All-to-all data exchange, rank 4 data, equal sizes.

See also:

mp_alltoall_i

Definition at line 352 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_alltoall_i45	(	INTEGER(KIND=int_4),dimension(:, :, :, :),intent(in)	sb,
		INTEGER(KIND=int_4),dimension(:, :, :, :, :),intent(out)	rb,
		INTEGER,intent(in)	count,
		INTEGER,intent(in)	group
	)

All-to-all data exchange, rank-4 data to rank-5 data.

Note:

User must ensure size consistency.

See also:

mp_alltoall_i

Definition at line 394 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_alltoall_i54	(	INTEGER(KIND=int_4),dimension(:, :, :, :, :),intent(in)	sb,
		INTEGER(KIND=int_4),dimension(:, :, :, :),intent(out)	rb,
		INTEGER,intent(in)	count,
		INTEGER,intent(in)	group
	)

All-to-all data exchange, rank-5 data to rank-4 data.

Note:

User must ensure size consistency.

See also:

mp_alltoall_i

Definition at line 474 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_bcast_i	(	INTEGER(KIND=int_4)	msg,
		INTEGER	source,
		INTEGER	gid
	)

Broadcasts a datum to all processes.

MPI mapping

mpi_bcast

Parameters:

[in]	msg	Datum to broadcast
[in]	source	Processes which broadcasts
[in]	gid	Message passing environment identifier

Definition at line 666 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_bcast_i3	(	INTEGER(KIND=int_4),dimension( :, :, : )	msg,
		INTEGER	source,
		INTEGER	gid
	)

Broadcasts rank-3 data to all processes.

Parameters:

[in]

msg

Data to broadcast

See also:

mp_bcast_i1

Definition at line 759 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_bcast_im	(	INTEGER(KIND=int_4),dimension( :, : )	msg,
		INTEGER	source,
		INTEGER	gid
	)

Broadcasts rank-2 data to all processes.

Parameters:

[in]

msg

Data to broadcast

See also:

mp_bcast_i1

Definition at line 728 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_bcast_iv	(	INTEGER(KIND=int_4),dimension( : )	msg,
		INTEGER	source,
		INTEGER	gid
	)

Broadcasts rank-1 data to all processes.

Parameters:

[in]

msg

Data to broadcast

See also:

mp_bcast_i1

Definition at line 697 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_deallocate_i	(	INTEGER(KIND=int_4),dimension(:),pointer	DATA,
		INTEGER,intent(out),optional	stat
	)

Deallocates special parallel memory.

Author:

UB

Parameters:

[in]

data

pointer to special memory to deallocate

Definition at line 2602 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_gather_i	(	INTEGER(KIND=int_4),intent(in)	msg,
		INTEGER(KIND=int_4),dimension( : ),intent(out)	msg_gather,
		INTEGER,intent(in)	root,
		INTEGER,intent(in)	gid
	)

Gathers a datum from all processes to one.

MPI mapping

mpi_gather

Parameters:

[in]	msg	Datum to send to root
[out]	msg_gather	Received data (on root)
[in]	root	Process which gathers the data
[in]	gid	Message passing environment identifier

Definition at line 1373 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_gather_im	(	INTEGER(KIND=int_4),dimension( :, : ),intent(in)	msg,
		INTEGER(KIND=int_4),dimension( :, : ),intent(out)	msg_gather,
		INTEGER,intent(in)	root,
		INTEGER,intent(in)	gid
	)

Gathers data from all processes to one.

Data length

All data (msg) is equal-sized

MPI mapping

mpi_gather

Parameters:

[in]

msg

Datum to send to root

See also:

mp_gather_i

Definition at line 1451 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_gather_iv	(	INTEGER(KIND=int_4),dimension( : ),intent(in)	msg,
		INTEGER(KIND=int_4),dimension( : ),intent(out)	msg_gather,
		INTEGER,intent(in)	root,
		INTEGER,intent(in)	gid
	)

Gathers data from all processes to one.

Data length

All data (msg) is equal-sized

MPI mapping

mpi_gather

Parameters:

[in]

msg

Datum to send to root

See also:

mp_gather_i

Definition at line 1412 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_gatherv_iv	(	INTEGER(KIND=int_4),dimension(:),intent(in)	sendbuf,
		INTEGER(KIND=int_4),dimension(:),intent(out)	recvbuf,
		INTEGER,dimension(:),intent(in)	recvcounts,
		INTEGER,dimension(:),intent(in)	displs,
		INTEGER,intent(in)	root,
		INTEGER,intent(in)	comm
	)

Gathers data from all processes to one.

Data length

Data can have different lengths

Offsets

Offsets start at 0

MPI mapping

mpi_gather

Parameters:

[in]	sendbuf	Data to send to root
[out]	recvbuf	Received data (on root)
[in]	recvcounts	Sizes of data received from processes
[in]	displs	Offsets of data received from processes
[in]	root	Process which gathers the data
[in]	comm	Message passing environment identifier

Definition at line 1496 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_irecv_im2	(	INTEGER(KIND=int_4),dimension(:, :),pointer	msgout,
		INTEGER,intent(in)	source,
		INTEGER,intent(in)	comm,
		INTEGER,intent(out)	request,
		INTEGER,intent(in),optional	tag
	)

Non-blocking receive of matrix data.

Note:

The argument must be a pointer to be sure that we do not get temporaries. They must point to contiguous memory.

Author:

fawzi

History

2009-11-25 [UB] Made type-generic for templates

See also:

mp_isendrecv_iv

mp_irecv_iv

Definition at line 2307 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_irecv_im3	(	INTEGER(KIND=int_4),dimension(:, :, :),pointer	msgout,
		INTEGER,intent(in)	source,
		INTEGER,intent(in)	comm,
		INTEGER,intent(out)	request,
		INTEGER,intent(in),optional	tag
	)

Non-blocking send of rank-3 data.

Note:

The argument must be a pointer to be sure that we do not get temporaries. They must point to contiguous memory.

Author:

fawzi

History

9.2008 added _rm3 subroutine [Iain Bethune] (c) The Numerical Algorithms Group (NAG) Ltd, 2008 on behalf of the HECToR project 2009-11-25 [UB] Made type-generic for templates

See also:

mp_isendrecv_iv

mp_irecv_iv

Definition at line 2365 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_irecv_iv	(	INTEGER(KIND=int_4),dimension(:),pointer	msgout,
		INTEGER,intent(in)	source,
		INTEGER,intent(in)	comm,
		INTEGER,intent(out)	request,
		INTEGER,intent(in),optional	tag
	)

Non-blocking receive of vector data.

Note:

The argument must be a pointer to be sure that we do not get temporaries. They must point to contiguous memory.

History

08.2003 created [f&j] 2009-11-25 [UB] Made type-generic for templates

See also:

mp_isendrecv_iv

Definition at line 2251 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_isend_im2	(	INTEGER(KIND=int_4),dimension(:, :),pointer	msgin,
		INTEGER,intent(in)	dest,
		INTEGER,intent(in)	comm,
		INTEGER,intent(out)	request,
		INTEGER,intent(in),optional	tag
	)

Non-blocking send of matrix data.

Note:

The argument must be a pointer to be sure that we do not get temporaries. They must point to contiguous memory.

Author:

fawzi

History

2009-11-25 [UB] Made type-generic for templates

See also:

mp_isendrecv_iv

mp_isend_iv

Definition at line 2133 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_isend_im3	(	INTEGER(KIND=int_4),dimension(:, :, :),pointer	msgin,
		INTEGER,intent(in)	dest,
		INTEGER,intent(in)	comm,
		INTEGER,intent(out)	request,
		INTEGER,intent(in),optional	tag
	)

Non-blocking send of rank-3 data.

Note:

The argument must be a pointer to be sure that we do not get temporaries. They must point to contiguous memory.

Author:

fawzi

History

9.2008 added _rm3 subroutine [Iain Bethune] (c) The Numerical Algorithms Group (NAG) Ltd, 2008 on behalf of the HECToR project 2009-11-25 [UB] Made type-generic for templates

See also:

mp_isendrecv_iv

mp_isend_iv

Definition at line 2192 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_isend_iv	(	INTEGER(KIND=int_4),dimension(:),pointer	msgin,
		INTEGER,intent(in)	dest,
		INTEGER,intent(in)	comm,
		INTEGER,intent(out)	request,
		INTEGER,intent(in),optional	tag
	)

Non-blocking send of vector data.

Note:

The argument must be a pointer to be sure that we do not get temporaries. They must point to contiguous memory.

History

08.2003 created [f&j]

See also:

mp_isendrecv_iv

Definition at line 2076 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_isendrecv_im2	(	INTEGER(KIND=int_4),dimension(:, :),pointer	msgin,
		INTEGER,intent(in)	dest,
		INTEGER(KIND=int_4),dimension(:, :),pointer	msgout,
		INTEGER,intent(in)	source,
		INTEGER,intent(in)	comm,
		INTEGER,intent(out)	send_request,
		INTEGER,intent(out)	recv_request,
		INTEGER,intent(in),optional	tag
	)

Non-blocking send and receieve of a matrix.

Note:

The arguments must be pointers to be sure that we do not get temporaries. They must point to contiguous memory.

History

08.2003 created [f&j]

See also:

mp_isendrecv_iv

Definition at line 2002 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_isendrecv_iv	(	INTEGER(KIND=int_4),dimension(:),pointer	msgin,
		INTEGER,intent(in)	dest,
		INTEGER(KIND=int_4),dimension(:),pointer	msgout,
		INTEGER,intent(in)	source,
		INTEGER,intent(in)	comm,
		INTEGER,intent(out)	send_request,
		INTEGER,intent(out)	recv_request,
		INTEGER,intent(in),optional	tag
	)

Non-blocking send and receieve of a vector.

Implementation

Calls mpi_isend and mpi_irecv.

Note:

The arguments must be pointers to be sure that we do not get temporaries. They must point to contiguous memory.

History

11.2004 created [Joost VandeVondele]

Parameters:

[in]	msgin	Vector data to send
[in]	dest	Which process to send to
[out]	msgout	Receive data into this pointer
[in]	source	Process to receive from
[in]	comm	Message passing environment identifier
[out]	send_request	Request handle for the send
[out]	recv_request	Request handle for the receive
[in]	tag	(optional) tag to differentiate requests

Definition at line 1932 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_max_i	(	INTEGER(KIND=int_4),intent(inout)	msg,
		INTEGER,intent(in)	gid
	)

Finds the maximum of a datum with the result left on all processes.

MPI mapping

mpi_allreduce

Parameters:

[in,out]	msg	Find maximum among these data (input) and maximum (output)
[in]	gid	Message passing environment identifier

Definition at line 1181 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_max_iv	(	INTEGER(KIND=int_4),dimension( : ),intent(inout)	msg,
		INTEGER,intent(in)	gid
	)

Finds the element-wise maximum of a vector with the result left on all processes.

Parameters:

[in,out]

msg

Find maximum among these data (input) and maximum (output)

See also:

mp_max_i

Definition at line 1216 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_min_i	(	INTEGER(KIND=int_4),intent(inout)	msg,
		INTEGER,intent(in)	gid
	)

Finds the minimum of a datum with the result left on all processes.

MPI mapping

mpi_allreduce

Parameters:

[in,out]	msg	Find minimum among these data (input) and maximum (output)
[in]	gid	Message passing environment identifier

Definition at line 1256 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_min_iv	(	INTEGER(KIND=int_4),dimension( : ),intent(inout)	msg,
		INTEGER,intent(in)	gid
	)

Finds the element-wise minimum of vector with the result left on all processes.

MPI mapping

mpi_allreduce

Parameters:

[in,out]

msg

Find minimum among these data (input) and maximum (output)

See also:

mp_min_i

Definition at line 1293 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_recv_i	(	INTEGER(KIND=int_4),intent(inout)	msg,
		INTEGER,intent(inout)	source,
		INTEGER,intent(inout)	tag,
		INTEGER,intent(in)	gid
	)

Receive one datum from another process.

MPI mapping

mpi_send

Parameters:

[in,out]	msg	Place received data into this variable
[in,out]	source	Process to receieve from
[in,out]	tag	Transfer identifier
[in]	gid	Message passing environment identifier

Definition at line 588 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_recv_iv	(	INTEGER(KIND=int_4),dimension( : ),intent(inout)	msg,
		INTEGER,intent(inout)	source,
		INTEGER,intent(inout)	tag,
		INTEGER,intent(in)	gid
	)

Receive rank-1 data from another process.

Parameters:

[in,out]

msg

Place receieved data into this rank-1 array

See also:

mp_recv_i

Definition at line 626 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_rma_get_i	(	TYPE(mp_window_type),intent(in)	window,
		INTEGER,intent(in)	remote,
		INTEGER,intent(in)	offset,
		INTEGER,intent(in)	len,
		INTEGER(KIND=int_4),dimension(:),pointer	storage
	)

Fetch access using RMA.

Author:

UB

Parameters:

[in]	window	RMA access window
[in]	remote	remote process
[in]	offset	offset in remote window (starting at 0)
[in]	len	amount of data to fetch
[out]	storage	integer storage of fetched data

Definition at line 2476 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_rma_put_i	(	TYPE(mp_window_type),intent(in)	window,
		INTEGER(KIND=int_4),dimension(:),pointer	storage,
		INTEGER,intent(in)	len,
		INTEGER,intent(in)	remote,
		INTEGER,intent(in)	offset
	)

Store access using RMA.

Author:

UB

Parameters:

[in]	window	RMA access window
[in]	storage	integer data to store
[in]	len	amount of data to store
[in]	remote	remote process
[in]	offset	offset in remote window (starting at 0)

Definition at line 2518 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_scatter_iv	(	INTEGER(KIND=int_4),dimension(:),intent(in)	msg_scatter,
		INTEGER(KIND=int_4),dimension( : ),intent(out)	msg,
		INTEGER,intent(in)	root,
		INTEGER,intent(in)	gid
	)

Scatters data from one processes to all others.

MPI mapping

mpi_scatter

Parameters:

[in]	msg_scatter	Data to scatter (for root process)
[out]	msg	Received data
[in]	root	Process which scatters data
[in]	gid	Message passing environment identifier

Definition at line 1334 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_send_i	(	INTEGER(KIND=int_4)	msg,
		INTEGER	dest,
		INTEGER	tag,
		INTEGER	gid
	)

Send one datum to another process.

MPI mapping

mpi_send

Parameters:

[in]	msg	Dum to send
[in]	dest	Destination process
[in]	tag	Transfer identifier
[in]	gid	Message passing environment identifier

Definition at line 519 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_send_iv	(	INTEGER(KIND=int_4),dimension( : )	msg,
		INTEGER	dest,
		INTEGER	tag,
		INTEGER	gid
	)

Send rank-1 data to another process.

Parameters:

[in]

msg

Rank-1 data to send

See also:

mp_send_i

Definition at line 552 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_sendrecv_im2	(	INTEGER(KIND=int_4),dimension( :, : ),intent(in)	msgin,
		INTEGER,intent(in)	dest,
		INTEGER(KIND=int_4),dimension( :, : ),intent(out)	msgout,
		INTEGER,intent(in)	source,
		INTEGER,intent(in)	comm
	)

Sends and receives matrix data.

See also:

mp_sendrecv_iv

Definition at line 1829 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_sendrecv_im3	(	INTEGER(KIND=int_4),dimension( :, :, : ),intent(in)	msgin,
		INTEGER,intent(in)	dest,
		INTEGER(KIND=int_4),dimension( :, :, : ),intent(out)	msgout,
		INTEGER,intent(in)	source,
		INTEGER,intent(in)	comm
	)

Sends and receives rank-3 data.

See also:

mp_sendrecv_iv

Definition at line 1874 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_sendrecv_iv	(	INTEGER(KIND=int_4),dimension( : ),intent(in)	msgin,
		INTEGER,intent(in)	dest,
		INTEGER(KIND=int_4),dimension( : ),intent(out)	msgout,
		INTEGER,intent(in)	source,
		INTEGER,intent(in)	comm
	)

Sends and receives vector data.

Parameters:

[in]	msgin	Data to send
[in]	dest	Process to send data to
[out]	msgout	Received data
[in]	source	Process from which to receive
[in]	comm	Message passing environment identifier

Definition at line 1784 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_shift_i	(	INTEGER(KIND=int_4),dimension( : ),intent(inout)	msg,
		INTEGER,intent(in)	group,
		INTEGER,intent(in),optional	displ_in
	)

Shift around the data in msg.

Example

msg will be moved from rank to rank+displ_in (in a circular way)

Limitations

* displ_in will be 1 by default (others not tested) * the message array needs to be the same size on all processes

Parameters:

[in,out]	msg	Data to shift
[in]	group	message passing environment identifier
[in]	displ_in	displacements (?)

Definition at line 72 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_shift_im	(	INTEGER(KIND=int_4),dimension( :, : ),intent(inout)	msg,
		INTEGER,intent(in)	group,
		INTEGER,intent(in),optional	displ_in
	)

Shift around the data in msg.

Example

msg will be moved from rank to rank+displ_in (in a circular way)

Limitations

* displ_in will be 1 by default (others not tested) * the message array needs to be the same size on all processes

Parameters:

[in,out]	msg	Rank-2 data to shift
[in]	group	message passing environment identifier
[in]	displ_in	displacements (?)

Definition at line 12 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_sum_i	(	INTEGER(KIND=int_4),intent(inout)	msg,
		INTEGER,intent(in)	gid
	)

Sums a datum from all processes with result left on all processes.

MPI mapping

mpi_allreduce

Parameters:

[in,out]	msg	Datum to sum (input) and result (output)
[in]	gid	Message passing environment identifier

Definition at line 793 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_sum_im	(	INTEGER(KIND=int_4),dimension( :, : ),intent(inout)	msg,
		INTEGER,intent(in)	gid
	)

Element-wise sum of a rank-2 array on all processes.

See also:

mp_sum_i

Parameters:

[in]

msg

Matrix to sum and result

Definition at line 866 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_sum_im3	(	INTEGER(KIND=int_4),dimension( :, :, : ),intent(inout)	msg,
		INTEGER,intent(in)	gid
	)

Element-wise sum of a rank-3 array on all processes.

See also:

mp_sum_i

Parameters:

[in]

msg

Arary to sum and result

Definition at line 906 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_sum_im4	(	INTEGER(KIND=int_4),dimension( :, :, :, : ),intent(inout)	msg,
		INTEGER,intent(in)	gid
	)

Element-wise sum of a rank-4 array on all processes.

See also:

mp_sum_i

Parameters:

[in]

msg

Arary to sum and result

Definition at line 948 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_sum_im5	(	INTEGER(KIND=int_4),dimension( :, :, :, :, : ),intent(inout)	msg,
		INTEGER,intent(in)	gid
	)

Element-wise sum of a rank-5 array on all processes.

See also:

mp_sum_i

Parameters:

[in]

msg

Arary to sum and result

Definition at line 991 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_sum_im6	(	INTEGER(KIND=int_4),dimension( :, :, :, :, :, : ),intent(inout)	msg,
		INTEGER,intent(in)	gid
	)

Element-wise sum of a rank-6 array on all processes.

See also:

mp_sum_i

Parameters:

[in]

msg

Arary to sum and result

Definition at line 1035 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_sum_iv	(	INTEGER(KIND=int_4),dimension( : ),intent(inout)	msg,
		INTEGER,intent(in)	gid
	)

Element-wise sum of a rank-1 array on all processes.

See also:

mp_sum_i

Parameters:

[in,out]

msg

Vector to sum and result

Definition at line 827 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_sum_root_im	(	INTEGER(KIND=int_4),dimension( :, : ),intent(inout)	msg,
		INTEGER,intent(in)	root,
		INTEGER,intent(in)	gid
	)

Element-wise sum of data from all processes with result left only on one.

Parameters:

[in,out]

msg

Matrix to sum (input) and (only on process root) result (output)

See also:

mp_sum_root_iv

Definition at line 1132 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_sum_root_iv	(	INTEGER(KIND=int_4),dimension( : ),intent(inout)	msg,
		INTEGER,intent(in)	root,
		INTEGER,intent(in)	gid
	)

Element-wise sum of data from all processes with result left only on one.

MPI mapping

mpi_reduce

Parameters:

[in,out]	msg	Vector to sum (input) and (only on process root) result (output)
[in]	gid	Message passing environment identifier

Definition at line 1085 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_sum_scatter_iv	(	INTEGER(KIND=int_4),dimension( : ),intent(in)	msgout,
		INTEGER(KIND=int_4),dimension( : ),intent(out)	msgin,
		INTEGER,dimension( : ),intent(in)	rcount,
		INTEGER,intent(in)	gid
	)

Sums a vector and partitions the result among processes.

Parameters:

[in]	msgout	Data to sum
[out]	msgin	Received portion of summed data
[in]	rcount	Partition sizes of the summed data for every process
[in]	gid	Message passing environment identifier

Definition at line 1743 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function:

subroutine mp_win_create_i	(	TYPE(mp_window_type),intent(out)	window,
		INTEGER(KIND=int_4),dimension(:),pointer	range,
		INTEGER,intent(in),optional	len,
		INTEGER,intent(in)	gid
	)

Creates an MPI RMA window.

Author:

UB

Parameters:

[out]	window	created window id
[in]	range	window contents
[in]	len	(optional) window size
[in]	gid	message passing environment identifier

Definition at line 2420 of file message_i_passing.f90.

References dbcsr_machine::m_walltime(), and timings::timeset().

Here is the call graph for this function: