Computer Science 335
Parallel Processing and High Performance Computing
Fall 2021, Siena College
Lecture 11: MPI Wrapup
Date: Friday, October 15, 2021
Agenda
- Announcements
- Midterm exam will be October 25
- complete in class
- pencil and paper only
- closed book/notes/etc
- based heavily on labs and programming projects
- heavy emphasis on MPI
- MPI Wrapup and Recap
- MPI processes do not share context (memory) directly
- An MPI implementation could use shared memory to implement
message passing when processes are on the same system
- MPI processes can be run on different computers, in which case
some form of network communication is needed
- MPI Basics: MPI_Init/MPI_Finalize,
MPI_Comm_size/MPI_Comm_rank,
MPI_Get_processor_name, MPI_Wtime
- Standard point-to-point communicaton:
MPI_Send/MPI_Recv, message tags, the MPI_Status
structure
- Non-blocking point-to-point communication:
MPI_Isend/MPI_Irecv,
MPI_Wait/MPI_Waitall/MPI_Waitany,
the MPI_Request structure
- Collective communication:
MPI_Reduce/MPI_Allreduce, MPI_Barrier,
MPI_Scan, MPI_Scatter/MPI_Gather
- Lab work