MPIProbe

MPI_Probe is a blocking MPI procedure used to test for the arrival of an incoming message and to obtain information about it without actually receiving the data. It waits until a message matching the specified source, tag, and communicator is available, then returns a status object containing details about the message. This preflight step is useful for planning the subsequent receive operation, such as sizing buffers or handling messages from varying sources and tags.

In C, the function is declared as int MPI_Probe(int source, int tag, MPI_Comm comm, MPI_Status *status). The source and tag can be specific values or MPI_ANY_SOURCE and MPI_ANY_TAG to match any incoming message from any source or with any tag within the given communicator. The call blocks until a matching message is available or an error occurs. When successful, the status object holds information about the message, including the actual source and tag.

After a successful probe, you typically determine how much data will be received and with what datatype. This is done by calling MPI_Get_count with the datatype you plan to receive, e.g., MPI_Get_count(&status, datatype, &count). The count indicates the number of elements of the given datatype in the incoming message. This enables you to allocate an appropriately sized receive buffer and then perform the corresponding MPI_Recv or MPI_Irecv operation. Note that Probe itself does not carry the data; it only prepares for the subsequent receive.

MPI_Iprobe is a non-blocking variant that checks for a message without blocking, returning a flag to indicate availability. MPI_Probe is commonly used to implement flexible, two-step communication patterns where the receiver decides how to handle an incoming message based on its source, tag, or size.