Parallel Programming, MPI Lecture 2

Transcription

1 Parallel Programming, MPI Lecture 2 Ehsan Nedaaee Oskoee 1 1 Department of Physics IASBS IPM Grid and HPC workshop IV, 2011

2 Outline 1 Introduction and Review The Von Neumann Computer Kinds of Parallel Machine Distributed memory parallel machines Shared memory parallel machine 2 Methodical Design Partitioning Domain Decomposition Functional Decomposition 3 An Introduction to MPI 4 Point-to-Point Communication Blocking PTP Communication

3 Outline 1 Introduction and Review The Von Neumann Computer Kinds of Parallel Machine Distributed memory parallel machines Shared memory parallel machine 2 Methodical Design Partitioning Domain Decomposition Functional Decomposition 3 An Introduction to MPI 4 Point-to-Point Communication Blocking PTP Communication

4 The Von Neumann Computer Figure: The Von Neumann Computer

5 Outline 1 Introduction and Review The Von Neumann Computer Kinds of Parallel Machine Distributed memory parallel machines Shared memory parallel machine 2 Methodical Design Partitioning Domain Decomposition Functional Decomposition 3 An Introduction to MPI 4 Point-to-Point Communication Blocking PTP Communication

6 Different type of parallel platforms:shared Memory Figure: The typical representation of a Shared Memory Parallel machine

7 Different type of parallel platforms:distributed Memory Figure: A typical representation of Distributed Memory Parallel machine

8 Outline 1 Introduction and Review The Von Neumann Computer Kinds of Parallel Machine Distributed memory parallel machines Shared memory parallel machine 2 Methodical Design Partitioning Domain Decomposition Functional Decomposition 3 An Introduction to MPI 4 Point-to-Point Communication Blocking PTP Communication

9 Methodical Design Partitioning Domain Decomposition Figure: Domain Decomposition 1 1 Designing and Building Parallel Programs (On-line book ), by Ian Foster

10 Methodical Design Partitioning Functional Decomposition Figure: Functional Decomposition 2 2 Designing and Building Parallel Programs (On-line book ), by Ian Foster

11 3 Methodical Design 3 Communication Number of tasks 8 8 = 64, Number of Communications 4 64 = 256.

12 3 Methodical Design 3 Communication Number of tasks 8 8 = 64, Number of Communications 4 64 = 256. Number of tasks 1 4 = 4, Number of Communications 4 4 = 16.

13 Methodical Design 4 Mapping 4 Designing and Building Parallel Programs (On-line book ), by Ian Foster

14 An Introduction to MPI Applications: Scalable Parallel Computers (SPCs) with distributed memory, Network Of Workstations (NOWs)

15 An Introduction to MPI Applications: Scalable Parallel Computers (SPCs) with distributed memory, Network Of Workstations (NOWs) Some Goals of MPI: Design an application programming interface, Allow efficient communication, Allow for implementations that can be used in a heterogeneous environment, Allow convenient C and Fortran 77 binding for the interface, Provide a reliable communication interface, Define an interface not too different from current practice, such as PVM, NX, etc.

16 An Introduction to MPI What is Included in MPI Point-to-point communication, Collective operations, Process groups, Communication domains, Process topologies, Environmental Management and inquiry, Profiling interface, Binding for Fortran 77 and C (Also for C++ and F90 in MPI-2)â, I/O functions (in MPI-2).â

17 An Introduction to MPI What is Included in MPI Point-to-point communication, Collective operations, Process groups, Communication domains, Process topologies, Environmental Management and inquiry, Profiling interface, Binding for Fortran 77 and C (Also for C++ and F90 in MPI-2)â, I/O functions (in MPI-2).â Versions of MPI Version 1.0 (was made in June 1994)â, Version 1.1 (was made in June 1995)â, Version 2.

18 An Introduction to MPI Procedure Specification The call uses but does not update an argument marked IN, The call may update an argument marked OUT, The call both uses and updates an argument marked INOUT.

19 An Introduction to MPI Procedure Specification The call uses but does not update an argument marked IN, The call may update an argument marked OUT, The call both uses and updates an argument marked INOUT. Types of MPI Calls Local, Non-local, Blocking, Non-blocking, Opaque Objects, Language Binding.

20 Outline 1 Introduction and Review The Von Neumann Computer Kinds of Parallel Machine Distributed memory parallel machines Shared memory parallel machine 2 Methodical Design Partitioning Domain Decomposition Functional Decomposition 3 An Introduction to MPI 4 Point-to-Point Communication Blocking PTP Communication

21 Point-to-Point Communication The Simplest Example main.for Program main implicit none include mpif.h integer ierr,rc call MPI_INIT(ierr) print*, HI There call MPI_FINALIZE(rc) End main.cpp #include <iostream.h> #include <mpi.h> int main(int argc, char **argv){ MPI_Init(&argc, &argv); cout«"hi There"; MPI_Finalize(); return 0; }

22 Point-to-Point Communication Compiling a Program > more hostfile Naft 2 Oil 1 > more hostfile HPCLAB ws01 ws02 ws03 Lamboot -v hostfile mpicc code_name.c -o code_exe_name mpicc code_name.cpp -o code_exe_name mpif77 code_name.for -o code_exe_name mpif90 code_name.f90 -o code_exe_name mpirun -v -np 9 code_exe_name mpirun N code_exe_name

23 Point-to-Point Communication A More Complex Program #include <iostream.h> #include <mpi.h> int main(int argc, char **argv) { int npes, myrank; MPI_Init(&argc, &argv); MPI_Comm_size(MPI_COMM_WORLD, &npes); MPI_Comm_rank(MPI_COMM_WORLD, &myrank); cout«"hi There, I am node "«myrank«" and the total worker which you are using now is:" «npes«endl; MPI_Finalize(); return 0; }

24 Point-to-Point Communication A More Complex Program Program size_rank implicit none include mpif.h integer ierr,npes,myrank call MPI_INIT(ierr) call MPI_COMM_SIZE( MPI_COMM_WORLD, npes, ierr ) call MPI_COMM_RANK( MPI_COMM_WORLD, myrank, ierr ) print*,"hi There, I am node ",myrank," and the total", *"number of workers which you are using now is: ",npes call MPI_FINALIZE(ierr) End

25 Point-to-Point Communication Blocking Send Operation MPI_SEND(buf, count, datatype, dest, tag, comm) IN buf initial address of send buffer IN count number of entries to send IN datatype datatype of each entry IN dest rank of destination IN tag message tag IN comm communicator C Version int MPI_Send(void* buf, int count, MPI_Datatype datatype, int dest, int tag, MPI_Comm, comm)

26 Point-to-Point Communication Blocking Send Operation MPI_SEND(buf, count, datatype, dest, tag, comm) IN buf initial address of send buffer IN count number of entries to send IN datatype datatype of each entry IN dest rank of destination IN tag message tag IN comm communicator Fortran version MPI_SEND(BUF, COUNT, DATATYPE, DEST, TAG, COMM, IERROR) <type> BUF(*) INTEGER COUNT, DATATYPE, DEST, TAG, COMM, IERROR

27 Point-to-Point Communication Blocking Receive Operation MPI_RECV(buf, count, datatype, source, tag, comm, status)â OUT buf initial address of received buffer IN count number of entries to receive IN datatype datatype of each entry IN source rank of source IN tag message tag IN comm communicator OUT status return status C Version int MPI_Recv(void* buf, int count, MPI_Datatype datatype, int source, int tag, MPI_Comm comm, MPI_Status *status);

28 Point-to-Point Communication Blocking Receive Operation MPI_RECV(buf, count, datatype, source, tag, comm, status)â OUT buf initial address of received buffer IN count number of entries to receive IN datatype datatype of each entry IN source rank of source IN tag message tag IN comm communicator OUT status return status Fortran Version MPI_RECV(BUF, COUNT, DATATYPE, SOURCE, TAG, COMM, STATUS, IERROR) <type> BUF(*) STATUS(MPI_STATUS_SIZE), IERROR INTEGER COUNT, DATATYPE, SOURCE, TAG, COMM,

29 Point-to-Point Communication Data Type MPI Data Type MPI_INTEGER MPI_REAL MPI_DOUBLE_PRECISION MPI_COMPLEX MPI_LOGICAL MPI_CHARACTER MPI_BYTE MPI_PACKED Fortran Data Type INTEGER REAL BOUBLE PRECISION COMPLEX LOGICAL CHARACTER(1)â

30 Point-to-Point Communication Data Type MPI Data Type MPI_CHAR MPI_SHORT MPI_INT MPI_LONG MPI_UNSIGNED_CHAR MPI_UNSIGNED_SHORT MPI_UNSIGNED MPI_UNSIGNED_LONG MPI_FLOAT MPI_DOUBLE MPI_LONG_DOUBLE MPI_BYTE MPI_PACKED C Data Type signed char signed short int signed int signed long int unsigned char unsigned short int unsigned int unsigned long int float double long double

31 Point-to-Point Communication A useful command MPI_GET_PROCESSOR_NAME(name, resultlen) OUT name A unique specifier for the current physical node OUT resultlen Length (in printable characters) of the result returned in name C Version int MPI_Get_processor_name(char* name, int* resultlen) Fortran Version MPI_GET_PROCESSOR_NAME(NAME, RESULTLEN, IERROR) CHARACTER*(*) NAME INTEGER RESULTLEN, IERROR

32 Point-to-Point Communication Blocking Send & Receive (Bsend_recd_1.cpp) int npes, myrank, namelen, i; char processor_name[mpi_max_processor_name]; char greeting[mpi_max_processor_name + 80]; MPI_Status status; MPI_Get_processor_name( processor_name, &namelen ); sprintf( greeting, "Hello, World, From process %d of %d on %s", myrank, npes, processor_name ); if (myrank == 0 ) { printf( "%s", greeting ); for (i = 1; i < npes; i++) { MPI_Recv( greeting, sizeof( greeting ), MPI_CHAR, i, 1, MPI_COMM_WORLD, &status); printf( "%s", greeting ); } } else { MPI_Send( greeting, strlen( greeting ) + 1, MPI_CHAR, 0, 1, MPI_COMM_WORLD); } } }

33 Point-to-Point Communication Blocking Send & Receive (Bsend_Recd_1.for) INTEGER ierr,npes,myrank,namelen,i INTEGER stat(mpi_status_size) CHARACTER*(MPI_MAX_PROCESSOR_NAME) processor_name CHARACTER (MPI_MAX_PROCESSOR_NAME + 80) greeting CHARACTER(1) numb(0:9) numb(0)="0" ; numb(1)="1"; numb(2)="2" ; numb(3)="3" numb(4)="4" ; numb(5)="5" ; numb(6)="6" ; numb(7)="7" numb(8)="8" ; numb(9)="9" call MPI_GET_PROCESSOR_NAME(processor_name,namelen,ierr) greeting = Hello Wold, From process //numb(myrank)// * of //numb(npes)// on //processor_name IF(myrank.EQ.0) THEN print*, greeting DO i = 1, npes -1 call MPI_RECV(greeting, len(greeting)+1,mpi_character,i,1, * MPI_COMM_WORLD,stat,ierr) print*,greeting END DO ELSE call MPI_SEND(greeting,len(greeting)+1,MPI_CHARACTER,0,1, * MPI_COMM_WORLD,ierr) ENDIF

34 Point-to-Point Communication Safety

35 Point-to-Point Communication Safety

36 Point-to-Point Communication Safety

37 Point-to-Point Communication Safety src = * :: MPI_ANY_SOURCE, tag = * :: MPI_ANY_TAG

38 Point-to-Point Communication Blocking Send Received example 2 Domain Decomposition

39 Point-to-Point Communication Blocking Send & Receive (Bsend_recd_2.cpp) for(j =0; j<= np+1; j++){ for(i=0; i<= np+1; i++)â a[i][j] = myrank*100+10*j+i;} left = myrank - 1; right= myrank + 1; if(myrank == 1 ) left = npes -1 ; if(myrank == (npes-1)) right = 1; if (myrank!= 0 ) { if(myrank%2==0){ MPI_Send(&a[1][0],np+2,MPI_INT,left,1,MPI_COMM_WORLD); MPI_Send(&a[np][0],np+2,MPI_INT,right,1,MPI_COMM_WORLD); MPI_Recv(&a[0][0],np+2,MPI_INT,left,1,MPI_COMM_WORLD,&stat); MPI_Recv(&a[np+1][0],np+2,MPI_INT,right,1,MPI_COMM_WORLD,&stat); } else { MPI_Recv(&a[np+1][0],np+2,MPI_INT,right,1,MPI_COMM_WORLD,&stat); MPI_Recv(&a[0][0],np+2,MPI_INT,left,1,MPI_COMM_WORLD,&stat); MPI_Send(&a[np][0],np+2,MPI_INT,right,1,MPI_COMM_WORLD); MPI_Send(&a[1][0],np+2,MPI_INT,left,1,MPI_COMM_WORLD); } }

40 Point-to-Point Communication Blocking Send & Receive (Bsend_Recd_2.for) IF(myrank.NE.0) THEN IF(MOD(myrank,2).EQ.0)THEN call MPI_SEND(a(0,1),np+2,MPI_INTEGER,left,1, * MPI_COMM_WORLD,ierr) call MPI_SEND(a(0,np),np+2,MPI_INTEGER,right,1, * MPI_COMM_WORLD,ierr) call MPI_RECV(a(0,0),np+2,MPI_INTEGER,left,1, * MPI_COMM_WORLD,stat,ierr) call MPI_RECV(a(0,np+1),np+2,MPI_INTEGER,right,1, * MPI_COMM_WORLD,stat,ierr) ELSE call MPI_RECV(a(0,np+1),np+2,MPI_INTEGER,right,1, * MPI_COMM_WORLD,stat,ierr) call MPI_RECV(a(0,0),np+2,MPI_INTEGER,left,1, * MPI_COMM_WORLD,stat,ierr) call MPI_SEND(a(0,np),np+2,MPI_INTEGER,right,1, * MPI_COMM_WORLD,ierr) call MPI_SEND(a(0,1),np+2,MPI_INTEGER,left,1, * MPI_COMM_WORLD,ierr) ENDIF ENDIF

41 The End That s All Folks