Blocking Communication
Here the rest of the code does not get executed until the the communication process is completed. Ex: MPI_Send, MPI_Receive
- Don’t have to worry about synchronisation.
- Less efficient
non-Blocking Communication
Here the rest of the code gets executed as the communication process is completes in the background. Ex: MPI_Isend, MPI_Ireceive
- Need additional functions like MPI_Wait() in order to ensure synchronization.
- Efficient use of computational resources.
MPI_Request send_request, recv_request;
// Non-Blocking Send
MPI_Isend(send_buffer, count, MPI_INT, dest, tag, MPI_COMM_WORLD, &send_request);
// Code execution continues immediately
// Non-Blocking Receive
MPI_Irecv(recv_buffer, count, MPI_INT, source, tag, MPI_COMM_WORLD, &recv_request);
// Code execution continues immediately
// Perform other useful work here while send/receive operations are in progress
// Ensure send and receive are complete before accessing the buffers
MPI_Wait(&send_request, MPI_STATUS_IGNORE);
MPI_Wait(&recv_request, MPI_STATUS_IGNORE);