Procedures

Aborts all processes in the communicator with the given error code

Calculate column distribution for a rank

Gathers integer values from all processes in the communicator

Blocks until all processes in the communicator have called barrier

Broadcasts integer data from root process to all processes in communicator

Broadcasts integer64 data from root process to all processes in communicator

Broadcasts integer64 data from root process to all processes in communicator

Broadcasts double precision data from root process to all processes in communicator

Broadcasts double precision data from root process to all processes in communicator

Broadcasts double precision array from root process to all processes in communicator

Broadcasts double precision array from root process to all processes in communicator

Broadcasts single-precision data from root process to all processes in communicator

Broadcasts single-precision data from root process to all processes in communicator

Broadcasts single-precision array from root process to all processes in communicator

Broadcasts single-precision array from root process to all processes in communicator

Frees the MPI communicator resources

Non-blocking probe for incoming messages

Initiates a non-blocking receive operation. The request must be waited on using wait() or test() before the buffer can be used.

Initiates a non-blocking receive operation. The request must be waited on using wait() or test() before the buffer can be used.

Initiates a non-blocking receive operation of an integer64

Initiates a non-blocking receive operation. The request must be waited on using wait() or test() before the buffer can be used.

Initiates a non-blocking receive operation of an integer64 array

Initiates a non-blocking receive operation. The request must be waited on using wait() or test() before the buffer can be used.

Non-blocking receive of a logical value

Non-blocking receive of a 2D allocatable double precision real array

Non-blocking receive with explicit count (for device pointers in host_data blocks)

Non-blocking receive with explicit count (for device pointers in host_data blocks)

Non-blocking receive of a single single-precision real

Non-blocking receive of a single-precision real array

Non-blocking receive of a 2D allocatable single-precision real array

Non-blocking receive with explicit count for single-precision (for device pointers in host_data blocks)

Non-blocking receive with explicit count for single-precision (for device pointers in host_data blocks)

Initiates a non-blocking send operation. The request must be waited on using wait() or test() before the buffer can be reused.

Initiates a non-blocking send operation. The request must be waited on using wait() or test() before the buffer can be reused.

Initiates a non-blocking send operation of an integer64

Initiates a non-blocking send operation. The request must be waited on using wait() or test() before the buffer can be reused.

Initiates a non-blocking send operation of an integer64 array

Initiates a non-blocking send operation. The request must be waited on using wait() or test() before the buffer can be reused.

Non-blocking send of a logical value

Initiates a non-blocking send operation. The request must be waited on using wait() or test() before the buffer can be reused.

Initiates a non-blocking send operation. The request must be waited on using wait() or test() before the buffer can be reused.

Non-blocking send of a 2D double precision real array

Non-blocking send of a contiguous 2D double-precision array with explicit count — no dim-prefix protocol. Pairs with comm_recv_real_dp_array_2d_n on the receive side. The caller must keep data valid until wait/waitall on the returned request completes.

Non-blocking send of a contiguous 2D dp array with explicit count. No dim-prefix protocol; pairs with comm_recv_real_dp_array_2d_n on the receive side.

Non-blocking send with explicit count (for device pointers in host_data blocks)

Non-blocking send with explicit count (for device pointers in host_data blocks)

Non-blocking send of a single single-precision real

Non-blocking send of a single-precision real array

Non-blocking send of a 2D single-precision real array

Non-blocking send with explicit count for single-precision (for device pointers in host_data blocks)

Non-blocking send with explicit count for single-precision (for device pointers in host_data blocks)

Returns true if the calling process has rank 0

Blocking probe for incoming messages. Returns once a message matching (source, tag) is queued at the receiver — caller reads status%MPI_SOURCE and status%MPI_TAG to decide who’s talking and what kind of message it is, then issues the matching recv. Use MPI_ANY_SOURCE / MPI_ANY_TAG to dispatch on whichever rank speaks first.

Blocking probe for incoming messages — caller reads status%MPI_SOURCE / status%MPI_TAG to dispatch the matching recv. Mirrors the f08 backend’s comm_probe.

Returns the 0-indexed rank of the calling process

Blocking receive of an integer from specified source. Use MPI_ANY_SOURCE or MPI_ANY_TAG for wildcards.

Blocking receive of an integer64 from specified source. Use MPI_ANY_SOURCE or MPI_ANY_TAG for wildcards.

Blocking receive of an integer64 from specified source

Blocking receive of an integer64 array from specified source. Array is automatically allocated to the correct size.

Blocking receive of an integer64 array from specified source

Blocking receive into a pre-allocated int64 array.

Blocking receive into a pre-allocated int64 array.

Blocking receive of an integer array from specified source.

Blocking recv of a contiguous 2D int32 array with explicit count.

Blocking recv of a contiguous 2D int32 array with explicit count.

Blocking receive into a pre-allocated int32 array.

Blocking receive into a pre-allocated int32 array.

Blocking receive of a logical value from specified source

Blocking receive of a single double precision real from specified source.

Blocking receive of a double precision real array from specified source.

Blocking receive of a 2D allocatable double precision real array

Receive 2D real(dp) array (must be pre-allocated by receiver)

Blocking receive of a contiguous 2D double-precision array with explicit count. No dim-prefix protocol — the caller has already shaped data. Mirrors comm_send_real_dp_array_2d_n.

Blocking recv of a contiguous 2D dp array with explicit count.

Blocking receive into a pre-allocated double-precision array.

Blocking receive into a pre-allocated double-precision array.

Blocking receive of a single single-precision real from specified source.

Blocking receive of a single-precision real array from specified source.

Blocking receive of a 2D allocatable single-precision real array

Receive 2D real(sp) array (must be pre-allocated by receiver)

Blocking receive into a pre-allocated single-precision array.

Blocking receive into a pre-allocated single-precision array.

Blocking send of an integer to specified destination

Blocking send of an integer64 to specified destination

Blocking send of an integer64 to specified destination

Blocking send of an integer64 array to specified destination

Blocking send of an integer64 array to specified destination

Blocking send of an integer array to specified destination

Blocking send of a contiguous 2D int32 array using an explicit count, no dim-prefix. Pairs with comm_recv_integer_array_2d_n.

Blocking send of a contiguous 2D int32 array with explicit count.

Blocking send of a logical value to specified destination

Blocking send of a single double precision real to specified destination

Blocking send of a double precision real array to specified destination

Blocking send of a 2D double precision real array to specified destination

Blocking send of a contiguous 2D double-precision array using an explicit count. Unlike comm_send_real_dp_array_2d, this variant does NOT prefix the message with the dimensions — caller and receiver agree on the shape via the protocol. The receiver uses comm_recv_real_dp_array_2d_n (or any of the _array_n overloads) with the same count.

Blocking send of a contiguous 2D dp array using an explicit count. No dim-prefix protocol — caller and receiver agree on shape externally.

Blocking send of a single single-precision real to specified destination

Blocking send of a single-precision real array to specified destination

Blocking send of a 2D single-precision real array to specified destination

Partitions the communicator into disjoint subgroups based on color. Processes with the same color end up in the same new communicator.

Creates a new communicator containing only processes that share memory with each other (typically processes on the same node)

Internal helper function that wraps an MPI_Comm into a comm_t object and caches rank and size information

Creates an invalid/null communicator object that can be used for initialization or to represent absence of a communicator.

Create MPI window for RMA operations

Create dynamic MPI window

Creates a new communicator that duplicates MPI_COMM_WORLD. This is the standard way to obtain a communicator for application use.

Destroy a distributed array

Query distribution for a rank

Global sum for double precision array (DDI_GSUMF equivalent)

Global sum for integer array (DDI_GSUMI equivalent)

Get number of processes and current rank (DDI_NPROC equivalent)

Synchronization barrier (DDI_SYNC equivalent)

Finalize the distributed arrays module

Initialize the distributed arrays module

Set communicator for convenience wrappers

Finalize dynamic load balancing

Initialize dynamic load balancing

Get next work unit atomically

Reset the load balancing counter to zero

Find a free registry slot

Get local offset in owner’s window for a given (row, col)

Find which rank owns a given column

Retrieves the name of the processor

Returns the number of processes in the communicator

Finalize MPI environment

Finalize MPI environment

Initialize MPI environment with optional threading support

Initialize MPI environment with optional threading support

Query the current MPI thread support level

Query the current MPI thread support level

Frees the MPI request resources

Retrieves the underlying MPI_Request handle

Checks if the request is null/invalid

Blocks until the operation associated with the request completes. The request is freed after completion.

Blocks until all operations in the request array complete. All requests are freed after completion.

Convert legacy integer array status to MPI_Status type