Basic MPI Programming Spring 2005

1
Basic MPI ProgrammingSpring 2005

• By Yaohang Li, Ph.D.
• Department of Computer Science
• North Carolina AT State University
• yaohang_at_ncat.edu

2
Review

• Last Class
• Definition of Parallel Computing System
• Hardware Characteristics
• Kinds of Processors
• Types of Memory Organization
• Flow of Control
• Interconnection Network
• Software Characteristics of Scientific Computing
  Applications
• Computation-bounded
• Communication-bounded
• Computation and Communication-bounded
• MPI
• Compiling and Running MPI programs
• MPI program Structure
• Primary elements
• Error handling
• Finding out the environment
• This Class
• MPI Programming

3
Better Hello (C)

• include "mpi.h"
• include ltstdio.hgt
• int main( int argc, char argv )
• int rank, size
• MPI_Init( argc, argv )
• MPI_Comm_rank( MPI_COMM_WORLD, rank )
• MPI_Comm_size( MPI_COMM_WORLD, size )
• printf( "I am d of d\n", rank, size )
• MPI_Finalize()
• return 0

4
Better Hello (Fortran)

• program main
• use MPI
• integer ierr, rank, size
• call MPI_INIT( ierr )
• call MPI_COMM_RANK( MPI_COMM_WORLD, rank, ierr )
• call MPI_COMM_SIZE( MPI_COMM_WORLD, size, ierr )
• print , 'I am ', rank, ' of ', size
• call MPI_FINALIZE( ierr )
• end

5
Communicators

• Communicators
• A parameter for most MPI calls
• A collection of processors working for a parallel
  job
• MPI_COMM_WORLD is defined in the MPI include file
  as all the processors in your program
• Can create subsets of MPI_COMM_WORLD
• Processors within a communicator are assigned
  numbers 0 to n-1

6
MPI Datatype

• Data Types
• When sending a message, it is given a data type
• Predefined types correspond to normal types
• MPI_REAL, MPI_FLOAT (Fortran Real and C float)
• MPI_DOUBLE_PRECISION, MPI_DOUBLE (Fortran double
  precision and C double)
• MPI_INTEGER and MPI_INT (Fortran and C integer)
• Can create user-defined type

7
Point-to-Point Communication

• Basic communication in message passing libraries
• Data values are transferred from one process to
  another
• One process sends the data
• Another receives the data

8
MPI_Send and MPI_Recv

• int MPI_Send(void buf, int count, MPI_Datatype
  datatype, int dest, int tag, MPI_Comm comm)
• Input
• buf - initial address of send buffer (choice)
• count - number of elements in send buffer
  (nonnegative integer)
• datatype - datatype of each send buffer element
  (handle)
• dest - rank of destination (integer)
• tag - message tag (integer)
• comm - communicator (handle)
• int MPI_Recv(void buf, int count, MPI_Datatype
  datatype, int source, int tag, MPI_Comm comm,
  MPI_Status status)
• Output
• buf - initial address of receive buffer
• status - status object, provides information
  about message received status is a structure of
  type MPI_Status, the element status.MPI_SOURCE is
  the source of the message received, and the
  element status.MPI_TAG is the tag value.
• Input
• count - maximum number of elements in receive
  buffer (integer)
• datatype - datatype of each receive buffer
  element (handle)
• source - rank of source (integer)
• tag - message tag (integer)
• comm - communicator (handle)

9
A Simple Send and Receive Program
tag1234 source0 destination1
count1 if(myid source)
buffer5678 MPI_Send(buffer,count,MPI_INT,desti
nation,tag,MPI_COMM_WORLD)
printf("processor d sent d\n",myid,buffer)
if(myid destination) MPI_Recv(buffer,c
ount,MPI_INT,source,tag,MPI_COMM_WORLD,status)
printf("processor d got
d\n",myid,buffer) MPI_Finalize()

• include ltstdio.hgt
• include "mpi.h"
• /
• This is a simple send/receive program in MPI
• /
• int main(argc,argv)
• int argc
• char argv
• int myid
• int tag,source,destination,count
• int buffer
• MPI_Status status
• MPI_Init(argc,argv)
• MPI_Comm_rank(MPI_COMM_WORLD,myid)

10
MPI is Simple

• Many parallel programs can be written using just
  these six functions, only two of which are
  non-trivial
• MPI_INIT
• MPI_FINALIZE
• MPI_COMM_SIZE
• MPI_COMM_RANK
• MPI_SEND
• MPI_RECV
• Point-to-point (send/recv) isnt the only way...

11
Introduction to Collective Operations in MPI

• Collective operations are called by all processes
  in a communicator.
• MPI_BCAST distributes data from one process (the
  root) to all others in a communicator.
• MPI_REDUCE combines data from all processes in
  communicator and returns it to one process.
• In many numerical algorithms, SEND/RECEIVE can be
  replaced by BCAST/REDUCE, improving both
  simplicity and efficiency.

12
MPI_BCAST

• MPI_BCAST( buffer, count, datatype, root, comm )
• INOUT buffer starting address of buffer
  (choice)
• IN count number of entries in buffer (integer)
  
• IN datatype data type of buffer (handle)
• IN root rank of broadcast root (integer)
• IN comm communicator (handle)
• C prototype
• int MPI_Bcast(void buffer, int count,
  MPI_Datatype datatype, int root, MPI_Comm comm)
• Fortran prototype
• MPI_BCAST(BUFFER, COUNT, DATATYPE, ROOT, COMM,
  IERROR)
• lttypegt BUFFER()
• INTEGER COUNT, DATATYPE, ROOT, COMM, IERROR

13
MPI_REDUCE

• MPI_REDUCE( sendbuf, recvbuf, count, datatype,
  op, root, comm)
• IN sendbuf address of send buffer (choice)
• OUT recvbuf address of receive buffer (choice,
  significant only at root)
• IN count number of elements in send buffer
  (integer)
• IN datatype data type of elements of send
  buffer (handle)
• IN op reduce operation (handle)
• Predefined operations including MPI_MAX, MPI_MIN
  and MPI_SUM
• IN root rank of root process (integer)
• IN comm communicator (handle)
• C Prototype
• int MPI_Reduce(void sendbuf, void recvbuf, int
  count, MPI_Datatype datatype, MPI_Op op, into
  root, MPI_Comm comm)
• Fortran Prototype
• MPI_REDUCE(SENDBUF, RECVBUF, COUNT, DATATYPE, OP,
  ROOT, COMM, IERROR)
• lttypegt SENDBUF(), RECVBUF()
• INTEGER COUNT, DATATYPE, OP, ROOT, COMM, IERROR

14
Example PI in C -1

• / Example program to calculate the value of pi
  by integrating f(x) 4 / (1 x2). /
• include "mpi.h"
• include ltmath.hgt
• int main(int argc, char argv)
• int done 0, n, myid, numprocs, i, rcdouble
  PI25DT 3.141592653589793238462643double mypi,
  pi, h, sum, x, aMPI_Init(argc,argv)MPI_Comm_
  size(MPI_COMM_WORLD,numprocs)MPI_Comm_rank(MPI_
  COMM_WORLD,myid) if (myid 0)
  printf("Enter the number of intervals ")
  scanf("d",n) MPI_Bcast(n, 1, MPI_INT,
  0, MPI_COMM_WORLD) if (n 0) break

15
Example PI in C - 2

• h 1.0 / (double) n sum 0.0 for (i
  myid 1 i lt n i numprocs) x h
  ((double)i - 0.5) sum 4.0 / (1.0 xx)
  mypi h sum MPI_Reduce(mypi, pi, 1,
  MPI_DOUBLE, MPI_SUM, 0,
  MPI_COMM_WORLD) if (myid 0) printf("pi
  is approximately .16f, Error is .16f\n",
  pi, fabs(pi - PI25DT))MPI_Finalize()
• return 0

16
Timing

• MPI Timing
• Performance Evaluation
• Debugging
• MPI Prototype
• MPI_WTIME()
• returns a floating-point number of seconds
• wall-clock time
• C double MPI_Wtime(void)
• Fortran DOUBLE PRECISION MPI_WTIME(

17
PI Program with Timing (I)

• include ltmpi.hgt
• include ltstdio.hgt
• include ltmath.hgt
• int main(int argc, char argv)
• int done 0, n, myid, numprocs, i, rc
• double PI25DT 3.141592653589793238462643
• double mypi, pi, h, sum, x, a
• double mytime
• char str100
• MPI_Init(argc,argv)
• MPI_Comm_size(MPI_COMM_WORLD,numprocs)
• MPI_Comm_rank(MPI_COMM_WORLD,myid)
• if (myid 0)
• printf("Enter the number of intervals ")
• scanf("d",n)

18
PI Program with Timing (II)

• MPI_Bcast(n, 1, MPI_INT, 0, MPI_COMM_WORLD)
• mytimeMPI_Wtime()
• h 1.0 / (double) n
• sum 0.0
• for (i myid 1 i lt n i numprocs)
• x h ((double)i - 0.5)
• sum 4.0 / (1.0 xx)
• mypi h sum
• mytimeMPI_Wtime()-mytime
• fprintf(stderr,"Computation time at process d
  f\n",myid,mytime)
• MPI_Reduce(mypi, pi, 1, MPI_DOUBLE, MPI_SUM,
  0, MPI_COMM_WORLD)
• if (myid 0)
• printf("pi is approximately .16f, Error is
  .16f\n", pi, fabs(pi - PI25DT))
• MPI_Finalize()
• return 0

19
Other MPI Functions

• MPI_GATHER
• MPI_SCATTER
• MPI_ALLgather
• MPI_ALLreduce
• Unblock communications
• Group Communications

20
Advantages of MPI Programming

• Universality
• MP model fits well on separate processors
  connected by fast/slow network
• Matches the hardware of most of todays parallel
  supercomputers as well as network of workstations
  (NOW)
• Expressivity
• MP has been found to be a useful and complete
  model in which to express parallel algorithms
• Ease of Debugging
• Debugging of parallel programs remains a
  challenging research area
• Debugging is easier in MPI paradigm than shared
  memory paradigm

21
Summary

• MPI Programming
• MPI Communicator
• MPI Point-to-Point Communication
• MPI Collective Communication
• MPI Timing
• Advantage of MPI

22
What I want you to do?

• Review Slides
• Read the UNIX handbook if you are not familiar
  with UNIX
• Read the introduction of MPI
• Work on your Assignment 1