MPI

Question 1

What is the function used to initialize the MPI execution environment?

Answer 1

MPI_Init

int MPI_Init(int *argc, char ***argv)

Question 2

What is the function used to finalise the MPI execution environment?

Answer 2

MPI_Finalize

int MPI_Finalize()

Question 3

What does MPI_Comm_size do?

Answer 3

Reports the number of MPI processes in the specified communicator

int MPI_Comm_size(MPI_Comm comm, int *size)

Question 4

What does MPI_Comm_rank do?

Answer 4

Reports the rank of the calling process in the specified communicator

int MPI_Comm_rank(MPI_Comm comm, int *rank)

Question 5

What is the range of ranks for MPI processes?

Answer 5

From 0 to size-1

Question 6

What is the predefined communicator that refers to all concurrent processes in an MPI program?

Answer 6

MPI_COMM_WORLD

Question 7

What are the two basic functions for point-to-point communication in MPI?

Answer 7

MPI_Send and MPI_Recv

Question 8

What is the purpose of the MPI_Send function?

Answer 8

Send a message to another process

int MPI_Send(void *buf, int count, MPI_Datatype datatype, int dest, int tag, MPI_Comm comm)

Question 9

What is the purpose of the MPI_Recv function?

Answer 9

Receive a message from another process

int MPI_Recv(void *buf, int count, MPI_Datatype datatype, int source, int tag, MPI_Comm comm, MPI_Status *status)

Question 10

List some MPI C data types.

Answer 10

• MPI_CHAR
• MPI_INT
• MPI_FLOAT
• MPI_DOUBLE
• etc

Question 11

What is the purpose of MPI_Bcast?

Answer 11

It is the function used to broadcast data from one process to all others in a communicator

int MPI_Bcast(void *buffer, int count, MPI_Datatype datatype, int root, MPI_Comm comm)

Question 12

What is the purpose of MPI_Scatter?

Answer 12

It is the function used to scatter data from one process to all others in a communicator.

int MPI_Scatter(const void *sendbuf, int sendcount, MPI_Datatype sendtype, void *recvbuf, int recvcount, MPI_Datatype recvtype, int root, MPI_Comm comm)

Question 13

What is purpose of MPI_Gather?

Answer 13

It is the function used to gather data from all processes in a communicator to one process.

int MPI_Gather(const void *sendbuf, int sendcount, MPI_Datatype sendtype, void *recvbuf, int recvcount, MPI_Datatype recvtype, int root, MPI_Comm comm)

Question 14

What is the purpose of MPI_Allgather?

Answer 14

It is the function used to gather data from all processes in a communicator to all processes.

int MPI_Allgather(const void *sendbuf, int sendcount, MPI_Datatype sendtype, void *recvbuf, int recvcount, MPI_Datatype recvtype, MPI_Comm comm)

Question 15

What are the two types of reduction functions in MPI?

Answer 15

• MPI_Reduce
• MPI_Allreduce

Question 16

What does MPI_Reduce do?

Answer 16

Carries out reduction and returns result to the specified process

int MPI_Reduce(const void *sendbuf, void *recvbuf, int count, MPI_Datatype type, MPI_Op op, int root, MPI_Comm comm)

Question 17

What does MPI_Allreduce do?

Answer 17

Carries out reduction and returns result to all processes

Question 18

True or False: MPI_COMM_SELF refers to all processes in an MPI program.

Answer 18

False

Question 19

What function is called to report the rank of this process?

Answer 19

MPI_Comm_rank

Question 20

What command is used to compile an MPI program in C?

Answer 20

mpicc

Question 21

What command is used to run an MPI process?

Answer 21

mpirun

Question 22

What are reduction variables in OpenMP?

Answer 22

Variables used to perform reduction operations across multiple processes

Communication overhead is greater if we return result to all processes

Question 23

What parameters are required for MPI_Reduce?

Answer 23

• sendbuf
• recvbuf
• count
• type
• op
• root
• comm

Question 24

What parameters are required for MPI_Allreduce?

Answer 24

• sendbuf
• recvbuf
• count
• type
• op
• comm

Question 25

What is blocking communication in MPI?

Answer 25

Communication where operations block until the message is received or safe to change buffers ## Footnote Examples include `MPI_Send` and `MPI_Recv`

Question 26

What is the behavior of `MPI_Send`?

Answer 26

Blocks until the message is received or it is safe to change the send buffer

Question 27

What is the behavior of `MPI_Recv`?

Answer 27

Blocks until the message is received

Question 28

How do non-blocking communications differ from blocking communications?

Answer 28

They allow overlapping computation and communication, avoiding strict matching of sends and receives ## Footnote Examples include `MPI_Isend` and `MPI_Irecv`

Question 29

What is the purpose of `MPI_Isend`?

Answer 29

To send a message non-blockingly ## Footnote `int MPI_Isend(void *buf, int count, MPI_Datatype datatype, int dest, int tag, MPI_Comm comm, MPI_Request *request)`

Question 30

What is the purpose of `MPI_Irecv`?

Answer 30

To receive a message non-blockingly ## Footnote `int MPI_Irecv(void *buf, int count, MPI_Datatype datatype, int source, int tag, MPI_Comm comm, MPI_Request *request)`

Question 31

What is the purpose of `MPI_Wait`?

Answer 31

To wait for a non-blocking request to complete ## Footnote `int MPI_Wait(MPI_Request *request, MPI_Status *status)`

Question 32

What is the purpose of `MPI_Waitall`?

Answer 32

To wait for all non-blocking requests to complete ## Footnote `int MPI_Waitall(int count, MPI_Request array_of_requests[], MPI_Status array_of_statuses[])`

Question 33

What are collectives in MPI?

Answer 33

Collective operations that require all processes in a communicator to reach the operation to avoid deadlocks

Question 34

What is Domain Decomposition in MPI?

Answer 34

A method to distribute work by dividing the computational domain among different MPI processes

Question 35

How is a PDE solver typically set up using MPI collectives?

Answer 35

1. Rank zero reads input parameters and broadcasts them 2. Rank zero reads initial conditions and scatters them 3. Rank zero gathers results and writes output

Question 36

What is the significance of using the same timestep in MPI PDE solvers?

Answer 36

All processes must use the same timestep for solution updates, which is determined by the minimum value across the computational domain

Question 37

What is the role of `MPI_Allreduce` with `MPI_MIN` in timestep calculation?

Answer 37

To determine the minimum timestep from the entire computational domain

Question 38

What are halos in the context of finite difference stencils?

Answer 38

Copies of rows/columns held by another MPI process to facilitate communication at domain boundaries

Question 39

How often do halos need to be updated?

Answer 39

Once per timestep

Question 40

What is a potential issue when using `MPI_Send` and `MPI_Recv` for halo communication?

Answer 40

It can incur significant overhead

Question 41

What is a non-blocking communication pattern for updating halos?

Answer 41

* `MPI_Isend` for sending halos * `MPI_Irecv` for receiving halos * `MPI_Waitall` to ensure completion