Parallel programming using MPI
1 like780 views
This document discusses collective communications in MPI (Message Passing Interface), including barriers, broadcast, gather/scatter, and reduction operations. It provides examples of how to use MPI_Barrier to synchronize processes, MPI_Bcast for one-to-all communication from a root process, MPI_Scatter and MPI_Gather for distributing and collecting data across processes, and a scatter example program. Collective operations require all processes in a communicator to call the routine and have no non-blocking versions.
Parallel programming using MPI
- 1. Parallel Programming using MPI Collective Communications Claudio Gheller CINECA [email protected] Collective Communications Collective Communications !"#$$%&'()*'#&+,'&-#.-'&/,),/0#%1,#2,10#(3++ !")..34,56,).. 10#(3++3+,'&,),(#$$%&'()*#0 Barrier Synchronization Broadcast Gather/Scatter Reduction (sum, max, prod, … ) 2 1
- 2. Collective Communications Characteristics All processes must call the collective routine No non-blocking collective communication No tags 7)23+*,)&4,$#+*,322'('3&*, (#$$%&'()*'#&,$#43 3 Collective Communications MPI_Barrier Stop processes until all processes within a communicator reach the barrier Fortran: CALL MPI_BARRIER(comm, ierr) C: int MPI_Barrier(MPI_Comm comm) 4 2
- 3. Collective Communications Barrier *: *; *< 8; 89 8= 8: 8< 89 8: 8; 8< 8= barrier barrier 89 8: 8; 8< 8= * 5 Collective Communications Broadcast (MPI_BCAST) One-to-all communication: same data sent from root process to all others in the communicator Fortran: INTEGER count, type, root, comm, ierr CALL MPI_BCAST(buf, count, type, root, comm, ierr) Buf array of type type C: int MPI_Bcast(void *buf, int count, MPI_Datatype, datatypem int root, MPI_Comm comm) All processes must specify same root, rank and comm 6 3
- 4. Collective Communications Broadcast PROGRAM broad_cast INCLUDE 'mpif.h' INTEGER ierr, myid, nproc, root INTEGER status(MPI_STATUS_SIZE) REAL A(2) 89 CALL MPI_INIT(ierr) CALL MPI_COMM_SIZE(MPI_COMM_WORLD, nproc, ierr) CALL MPI_COMM_RANK(MPI_COMM_WORLD, myid, ierr) ): root = 0 8: IF( myid .EQ. 0 ) THEN 89 ): a(1) = 2.0 a(2) = 4.0 ): END IF 8; CALL MPI_BCAST(a, 2, MPI_REAL, 0, ): MPI_COMM_WORLD, ierr) WRITE(6,*) myid, ': a(1)=', a(1), 'a(2)=', a(2) CALL MPI_FINALIZE(ierr) 8< END 7 Collective Communications Scatter / Gather 7()**30 >)*?30 sndbuf sndbuf sndbuf sndbuf sndbuf 89 89 ): ); )< )= 89 89 ): 8: 8: ); 8; 8; )< 8< 8< )= 89 ): 8: ); 8; )< 8< )= 89 89 ): ); )< )= rcvbuf rcvbuf rcvbuf rcvbuf rcvbuf 8 4
- 5. Collective Communications MPI_Scatter One-to-all communication: different data sent from root process to all others in the communicator Fortran: CALL MPI_SCATTER(sndbuf, sndcount, sndtype, rcvbuf, rcvcount, rcvtype, root, comm, ierr) Arguments definition are like other MPI subroutine sndcount is the number of elements sent to each process, not the size of sndbuf, that should be sndcount times the number of process in the communicator The sender arguments are meaningful only for root 9 Collective Communications 10 5
- 6. Collective Communications MPI_SCATTERV Usage – int MPI_Scatterv( void* sendbuf, /* in */ int* sendcounts, /* in */ int* displs, /* in */ MPI_Datatype sendtype, /* in */ void* recvbuf, /* in */ int recvcount, /* in */ MPI_Datatype recvtype, /* in */ int root, /* in */ MPI_Comm comm); /* in */ Description – Distributes individual messages from root to each process in communicator – Messages can have different sizes and displacements 11 Collective Communications 12 6
- 7. Collective Communications MPI_Gather One-to-all communication: different data collected by the root process, from all others processes in the communicator. Is the opposite of Scatter Fortran: CALL MPI_GATHER(sndbuf, sndcount, sndtype, rcvbuf, rcvcount, rcvtype, root, comm, ierr) Arguments definition are like other MPI subroutine rcvcount is the number of elements collected from each process, not the size of rcvbuf, that should be rcvcount times the number of process in the communicator The receiver arguments are meaningful only for root 13 Collective Communications 14 7
- 8. Collective Communications MPI_GATHERV Usage – int MPI_Gatherv( void* sendbuf, /* in */ int sendcount, /* in */ MPI_Datatype sendtype, /* in */ void* recvbuf, /* out */ int* recvcount, /* in */ int* displs, /* in */ MPI_Datatype recvtype, /* in */ int root, /* in */ MPI_Comm comm ); /* in */ Description – Collects individual messages from each process in communicator to the root process and store them in rank order – Messages can have different sizes and displacements 15 Collective Communications 16 8
- 9. Collective Communications Scatter example PROGRAM scatter INCLUDE 'mpif.h' INTEGER ierr, myid, nproc, nsnd, I, root INTEGER status(MPI_STATUS_SIZE) REAL A(16), B(2) CALL MPI_INIT(ierr) CALL MPI_COMM_SIZE(MPI_COMM_WORLD, nproc, ierr) CALL MPI_COMM_RANK(MPI_COMM_WORLD, myid, ierr) root = 0 IF( myid .eq. root ) THEN DO i = 1, 16 a(i) = REAL(i) END DO END IF nsnd = 2 CALL MPI_SCATTER(a, nsnd, MPI_REAL, b, nsnd, & MPI_REAL, root, MPI_COMM_WORLD, ierr) WRITE(6,*) myid, ': b(1)=', b(1), 'b(2)=', b(2) CALL MPI_FINALIZE(ierr) 17 Collective Communications Gather example PROGRAM gather INCLUDE 'mpif.h' INTEGER ierr, myid, nproc, nsnd, I, root INTEGER status(MPI_STATUS_SIZE) REAL A(16), B(2) CALL MPI_INIT(ierr) CALL MPI_COMM_SIZE(MPI_COMM_WORLD, nproc, ierr) CALL MPI_COMM_RANK(MPI_COMM_WORLD, myid, ierr) root = 0 b(1) = REAL( myid ) b(2) = REAL( myid ) nsnd = 2 CALL MPI_GATHER(b, nsnd, MPI_REAL, a, nsnd, & MPI_REAL, root MPI_COMM_WORLD, ierr) IF( myid .eq. root ) THEN DO i = 1, (nsnd*nproc) WRITE(6,*) myid, ': a(i)=', a(i) END DO END IF CALL MPI_FINALIZE(ierr) 18 9
- 10. Collective Communications MPI_Alltoall @#0*0)&A CALL MPI_ALLTOALL(sndbuf, sndcount, sndtype, rcvbuf, rcvcount, rcvtype, comm, ierr) 89 ): ); )< )= 89 ): 5: (: 4: 89 89 rcvbuf sndbuf 5: 5; 5< 5= ); 5; (; 4; 89 (: (; (< (= 89 )< 5< (< 4< 89 4: 4; 4< 4= 89 )= 5= (= 4= B306,%+32%.,*#,'$1.3$3&*,4)*),*0)&+1#+'*'#& 19 Collective Communications Reduction The reduction operation allow to: • Collect data from each process • Reduce the data to a single value • Store the result on the root processes • Store the result on all processes • Overlap of communication and computing 20 10
- 11. Collective Communications Reduce, Parallel Sum 89 ): 5: 89 7) 75 8: ); 5; 8: 7) 75 7)C):D);D)<D)= 8; )< 5< 75C5:D5;D5<D5= 8; 7) 75 8< )= 5= 8< 7) 75 E34%(*'#&,2%&(*'#&,F#0G+,F'*?,)00)6+ #*?30,#130)*'#&A,10#4%(*H,$'&H,$)IH,)&4H,JK, 21 Collective Communications MPI_REDUCE and MPI_ALLREDUCE Fortran: MPI_REDUCE( snd_buf, rcv_buf, count, type, op, root, comm, ierr) snd_buf input array of type type containing local values. rcv_buf output array of type type containing global results Count (INTEGER) number of element of snd_buf and rcv_buf type (INTEGER) MPI type of snd_buf and rcv_buf op (INTEGER) parallel operation to be performed root (INTEGER) MPI id of the process storing the result comm (INTEGER) communicator of processes involved in the operation ierr (INTEGER) output, error code (if ierr=0 no error occours) MPI_ALLREDUCE( snd_buf, rcv_buf, count, type, op, comm, ierr) The argument root is missing, the result is stored to all processes. 22 11
- 12. Collective Communications Predefined Reduction Operations MPI op Function MPI_MAX Maximum MPI_MIN Minimum MPI_SUM Sum MPI_PROD Product MPI_LAND Logical AND MPI_BAND Bitwise AND MPI_LOR Logical OR MPI_BOR Bitwise OR MPI_LXOR Logical exclusive OR MPI_BXOR Bitwise exclusive OR MPI_MAXLOC Maximum and location MPI_MINLOC Minimum and location 23 Collective Communications Reduce / 1 C: int MPI_Reduce(void * snd_buf, void * rcv_buf, int count, MPI_Datatype type, MPI_Op op, int root, MPI_Comm comm) int MPI_Allreduce(void * snd_buf, void * rcv_buf, int count, MPI_Datatype type, MPI_Op op, MPI_Comm comm) 24 12
- 13. Collective Communications Reduce, example PROGRAM reduce INCLUDE 'mpif.h' INTEGER ierr, myid, nproc, root INTEGER status(MPI_STATUS_SIZE) REAL A(2), res(2) CALL MPI_INIT(ierr) CALL MPI_COMM_SIZE(MPI_COMM_WORLD, nproc, ierr) CALL MPI_COMM_RANK(MPI_COMM_WORLD, myid, ierr) root = 0 a(1) = 2.0*myid a(2) = 4.0+myid CALL MPI_REDUCE(a, res, 2, MPI_REAL, MPI_SUM, root, & MPI_COMM_WORLD, ierr) IF( myid .EQ. 0 ) THEN WRITE(6,*) myid, ': res(1)=', res(1), 'res(2)=', res(2) END IF CALL MPI_FINALIZE(ierr) END 25 13