Upload
joey
View
28
Download
0
Embed Size (px)
DESCRIPTION
MPI Part III NPACI Parallel Computing Institute August 19 - 23, 2002 San Diego Supercomputer Center. Point to Point Communications in MPI. Basic operations of Point to Point (PtoP) communication and issues of deadlock Several steps are involved in the PtoP communication Sending process - PowerPoint PPT Presentation
Citation preview
1San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI Part III
NPACI Parallel Computing InstituteAugust 19 - 23, 2002
San Diego Supercomputer Center
2San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Point to Point Communications in MPI
• Basic operations of Point to Point (PtoP) communication and issues of deadlock
• Several steps are involved in the PtoP communication
• Sending process– data is copied to the user buffer by the user
– User calls one of the MPI send routines
– System copies the data from the user buffer to the system buffer
– System sends the data from the system buffer to the destination processor
3San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Point to Point Communications in MPI
• Receiving process– User calls one of the MPI receive subroutines– System receives the data from the source process, and
copies it to the system buffer– System copies the data from the system buffer to the
user buffer– User uses the data in the user buffer
4San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
sendbuf
Call send routine
Now sendbuf can be reused
Process 0 : User mode
Kernel mode
Copying data from sendbuf to systembuf
Send data from sysbuf to dest
dataProcess 1 : User mode Kernel mode
Call receive routinereceive data from src to systembuf
Copying data from sysbufto recvbuf
sysbuf
sysbuf
recvbuf
Now recvbuf contains valid data
5San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Unidirectional communication
• Blocking send and blocking receive
if (myrank == 0) then call MPI_Send(…)
elseif (myrank == 1) thencall MPI_Recv(….)
endif
• Non-blocking send and blocking receive
if (myrank == 0) thencall MPI_ISend(…)call MPI_Wait(…)
else if (myrank == 1) thencall MPI_Recv(….)
endif
6San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
• Blocking send and non-blocking recvif (myrank == 0 ) then
call MPI_Send(…..)elseif (myrank == 1) then
call MPI_Irecv (…)call MPI_Wait(…)
endif
• Non-blocking send and non-blocking recvif (myrank == 0 ) then
call MPI_Isend (…)call MPI_Wait (…)
elseif (myrank == 1) thencall MPI_Irecv (….)call MPI_Wait(..)
endif
Unidirectional communication
7San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Bidirectional communication
• Need to be careful about deadlock when two processes exchange data with each other
• Deadlock can occur due to incorrect order of send and recv or due to limited size of the system buffer
sendbuf
recvbuf
Rank 0 Rank 1
recvbuf
sendbuf
8San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Bidirectional communication• Case 1 : both processes call send first, then recv
if (myrank == 0 ) thencall MPI_Send(….)call MPI_Recv (…)
elseif (myrank == 1) thencall MPI_Send(….)call MPI_Recv(….)
endif
• No deadlock as long as system buffer is larger than send buffer
• Deadlock if system buffer is smaller than send buf
• If you replace MPI_Send with MPI_Isend and MPI_Wait, it is still the same
• Moral : there may be error in coding that only shows up for larger problem size
9San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Bidirectional communication
• Following is free from deadlock
if (myrank == 0 ) thencall MPI_Isend(….)call MPI_Recv (…)call MPI_Wait(…)
elseif (myrank == 1) thencall MPI_Isend(….)call MPI_Recv(….)call MPI_Wait(….)
endif
10San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Bidirectional communication
• Case 2 : both processes call recv first, then send if (myrank == 0 ) then
call MPI_Recv(….)call MPI_Send (…)
elseif (myrank == 1) thencall MPI_Recv(….)call MPI_Send(….)
endif
• The above will always lead to deadlock (even if you replace MPI_Send with MPI_Isend and MPI_Wait)
11San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Bidirectional communication
• The following code can be safely executed
if (myrank == 0 ) thencall MPI_Irecv(….)call MPI_Send (…)call MPI_Wait(…)
elseif (myrank == 1) thencall MPI_Irecv(….)call MPI_Send(….)call MPI_Wait(….)
endif
12San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Bidirectional communication
• Case 3 : one process call send and recv in this order, and the other calls in the opposite order
if (myrank == 0 ) thencall MPI_Send(….)call MPI_Recv(…)
elseif (myrank == 1) thencall MPI_Recv(….)call MPI_Send(….)
endif
• The above is always safe
• You can replace both send and recv on both processor with Isend and Irecv
13San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Scatter and Gather A Ap0
p1
p2
p3
p0
p1
p2
p3
A
A
A
broadcast
scatterA B C D A
B
C
Dgather
A
B
C
D
A B C D
A B C D
A B C D
A B C D
all gather
p0
p1
p2
p3
p0
p1
p2
p3
p0
p1
p2
p3
p0
p1
p2
p3
14San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Scatter Operation using MPI_Scatter
• Similar to Broadcast but sends a section of an array to each processors
A(0) A(1) A(2) . . ………. A(N-1)
P0 P1 P2 . . . Pn-1
Goes to processors:
Data in an array on root node:
15San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI_Scatter
• C – int MPI_Scatter(&sendbuf, sendcnts, sendtype, &recvbuf,
recvcnts, recvtype, root, comm );
• Fortran – MPI_Scatter(sendbuf,sendcnts,sendtype,
recvbuf,recvcnts,recvtype,root,comm,ierror)
• Parameters– sendbuf is an array of size (number processors*sendcnts)
– sendcnts number of elements sent to each processor
– recvcnts number of element(s) obtained from the root processor
– recvbuf contains element(s) obtained from the root processor, may be an array
16San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Scatter Operation using MPI_Scatter
• Scatter with Sendcnts = 2
A(0) A(2) A(4) . . . A(2N-2) A(1) A(3) A(5) . . . A(2N-1)
P0 P1 P2 . . . Pn-1
B(0) B(0) B(0) B(0)B(1) B(1) B(1) B(1)
Goes to processors:
Data in an array on root node:
17San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Gather Operation using MPI_Gather
• Used to collect data from all processors to the root, inverse of scatter
• Data is collected into an array on root processor
A(0) A(1) A(2) . . . A(N-1)
P0 P1 P2 . . . Pn-1
A0 A1 A2 . . . An-1
Data from variousProcessors:
Goes to an array on root node:
18San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI_Gather
• C – int MPI_Gather(&sendbuf,sendcnts, sendtype, &recvbuf,
recvcnts,recvtype,root, comm );
• Fortran – MPI_Gather(sendbuf,sendcnts,sendtype,
recvbuf,recvcnts,recvtype,root,comm,ierror)
• Parameters– sendcnts number of elements sent from each processor
– sendbuf is an array of size sendcnts
– recvcnts number of elements obtained from each processor
– recvbuf of size recvcnts*number of processors
19San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Code for Scatter and Gather
• A parallel program to scatter data using MPI_Scatter
• Each processor sums the data
• Use MPI_Gather to get the data back to the root processor
• Root processor prints the global data
• See attached Fortran and C code
20San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
module mpi!DEC$ NOFREEFORM include "mpif.h“!DEC$ FREEFORM
end module! This program shows how to use MPI_Scatter and MPI_Gather! Each processor gets different data from the root processor! by way of mpi_scatter. The data is summed and then sent back! to the root processor using MPI_Gather. The root processor! then prints the global sum.
module global integer numnodes,myid,mpi_err integer, parameter :: mpi_root=0end modulesubroutine init use mpi use global implicit none! do the mpi init stuff call MPI_INIT( mpi_err ) call MPI_COMM_SIZE( MPI_COMM_WORLD, numnodes, mpi_err ) call MPI_Comm_rank(MPI_COMM_WORLD, myid, mpi_err)
21San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
end subroutine initprogram test1 use mpi use global implicit none integer, allocatable :: myray(:),send_ray(:),back_ray(:) integer count integer size,mysize,i,k,j,total call init! each processor will get count elements from the root count=4 allocate(myray(count))! create the data to be sent on the root if(myid == mpi_root)then size=count*numnodes allocate(send_ray(0:size-1)) allocate(back_ray(0:numnodes-1)) do i=0,size-1 send_ray(i)= i enddo endif
22San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
call MPI_Scatter( send_ray, count, MPI_INTEGER, & myray, count, MPI_INTEGER, & mpi_root, MPI_COMM_WORLD,mpi_err)! each processor does a local sum total=sum(myray) write(*,*)"myid= ",myid," total= ",total! send the local sums back to the root call MPI_Gather( total, 1, MPI_INTEGER, & back_ray, 1, MPI_INTEGER, & mpi_root, MPI_COMM_WORLD,mpi_err)! the root prints the global sum if(myid == mpi_root)then write(*,*)"results from all processors= ",sum(back_ray) endif call mpi_finalize(mpi_err)
end program
23San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
#include <mpi.h>#include <stdio.h>#include <stdlib.h>/*! This program shows how to use MPI_Scatter and MPI_Gather! Each processor gets different data from the root processor! by way of mpi_scatter. The data is summed and then sent back! to the root processor using MPI_Gather. The root processor! then prints the global sum. *//* globals */int numnodes,myid,mpi_err;#define mpi_root 0/* end globals */void init_it(int *argc, char ***argv);void init_it(int *argc, char ***argv) { mpi_err = MPI_Init(argc,argv); mpi_err = MPI_Comm_size( MPI_COMM_WORLD, &numnodes ); mpi_err = MPI_Comm_rank(MPI_COMM_WORLD, &myid); }
24San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
int main(int argc,char *argv[]){ int *myray,*send_ray,*back_ray; int count; int size,mysize,i,k,j,total; init_it(&argc,&argv);/* each processor will get count elements from the root */ count=4; myray=(int*)malloc(count*sizeof(int));/* create the data to be sent on the root */ if(myid == mpi_root){ size=count*numnodes; send_ray=(int*)malloc(size*sizeof(int)); back_ray=(int*)malloc(numnodes*sizeof(int)); for(i=0;i<size;i++) send_ray[i]=i; }
/* send different data to each processor */
25San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
mpi_err = MPI_Scatter( send_ray, count, MPI_INT, myray, count, MPI_INT, mpi_root, MPI_COMM_WORLD);/* each processor does a local sum */ total=0; for(i=0;i<count;i++) total=total+myray[i]; printf("myid= %d total= %d\n ",myid,total);/* send the local sums back to the root */ mpi_err = MPI_Gather(&total, 1, MPI_INT, back_ray, 1, MPI_INT, mpi_root, MPI_COMM_WORLD);/* the root prints the global sum */ if(myid == mpi_root){ total=0; for(i=0;i<numnodes;i++) total=total+back_ray[i]; printf("results from all processors= %d \n ",total); } mpi_err = MPI_Finalize();}
26San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Output of previous fortran code on 4 procs
ultra:/work/majumdar/examples/mpi % bsub -q hpc -m ultra -I -n 4 ./a.out
Job <48051> is submitted to queue <hpc>.
<<Waiting for dispatch ...>>
<<Starting on ultra>>
myid= 1 total= 22
myid= 2 total= 38
myid= 3 total= 54
myid= 0 total= 6
results from all processors= 120
( 0 through 15 added up = (15) (15 + 1) /2 = 120)
27San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Global Sum with MPI_Reduce2d array spread across processors
A0+A1+A2 B0+B1+B2 C0+C1+C2NODE 0NODE 1NODE 2
X(0) X(1) X(2)
A0 B0 C0A1 B1 C1A2 B2 C2
NODE 0NODE 1NODE 2
X(0) X(1) X(2)
28San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
All Gather and All Reduce
• Gather and Reduce come in an "ALL" variation
• Results are returned to all processors
• The root parameter is missing from the call
• Similar to a gather or reduce followed by a broadcast
29San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Global Sum with MPI_AllReduce 2d array spread across processors
A0 B0 C0A1 B1 C1A2 B2 C2
X(0) X(1) X(2)
NODE 0 NODE 1 NODE 2
A0+A1+A2 B0+B1+B2 C0+C1+C2A0+A1+A2 B0+B1+B2 C0+C1+C2A0+A1+A2 B0+B1+B2 C0+C1+C2
X(0) X(1) X(2)
NODE 0 NODE 1 NODE 2
30San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
All to All communication with MPI_Alltoall
• Each processor sends and receives data to/from all others
• C – int MPI_Alltoall(&sendbuf,sendcnts, sendtype, &recvbuf,
recvcnts, recvtype, MPI_Comm);
• Fortran – call MPI_Alltoall(sendbuf,sendcnts,sendtype,
recvbuf,recvcnts,recvtype,comm,ierror)
31San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
b0 b1 b2 b3c0 c1 c2 c3d0 d1 d2 d3
a0 a1 a2 a3a1 b1 c1 d1a2 b2 c2 d2a3 b3 c3 d3
a0 b0 c0 d0All to All
32San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
All to All with MPI_Alltoall
• Parameters– sendcnts # of elements sent to each processor
– sendbuf is an array of size sendcnts
– recvcnts # of elements obtained from each processor
– recvbuf of size recvcnts
• Note that both send buffer and receive buffer must be an array of size of the number of processors
• See attached Fortran and C codes
33San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
module mpi!DEC$ NOFREEFORM include "mpif.h“!DEC$ FREEFORM end module! This program shows how to use MPI_Alltoall. Each processor! send/rec a different random number to/from other processors. module global integer numnodes,myid,mpi_err integer, parameter :: mpi_root=0end modulesubroutine init use mpi use global implicit none! do the mpi init stuff call MPI_INIT( mpi_err ) call MPI_COMM_SIZE( MPI_COMM_WORLD, numnodes, mpi_err ) call MPI_Comm_rank(MPI_COMM_WORLD, myid, mpi_err)end subroutine init
34San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
program test1 use mpi use global implicit none integer, allocatable :: scounts(:),rcounts(:) integer ssize,rsize,i,k,j real z call init ! counts and displacement arrays allocate(scounts(0:numnodes-1)) allocate(rcounts(0:numnodes-1)) call seed_random! find data to send do i=0,numnodes-1 call random_number(z) scounts(i)=nint(10.0*z)+1 Enddo write(*,*)"myid= ",myid," scounts= ",scounts
35San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
! send the data call MPI_alltoall( scounts,1,MPI_INTEGER, & rcounts,1,MPI_INTEGER, MPI_COMM_WORLD,mpi_err) write(*,*)"myid= ",myid," rcounts= ",rcounts call mpi_finalize(mpi_err)end program subroutine seed_random use global implicit none integer the_size,j integer, allocatable :: seed(:) real z call random_seed(size=the_size) ! how big is the intrisic seed? allocate(seed(the_size)) ! allocate space for seed do j=1,the_size ! create the seed seed(j)=abs(myid*10)+(j*myid*myid)+100 ! abs is generic enddo call random_seed(put=seed) ! assign the seed deallocate(seed)end subroutine
36San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
#include <mpi.h>#include <stdio.h>|#include <stdlib.h>/*! This program shows how to use MPI_Alltoall. Each processor! send/rec a different random number to/from other processors. *//* globals */int numnodes,myid,mpi_err;#define mpi_root 0/* end module */void init_it(int *argc, char ***argv);void seed_random(int id);void random_number(float *z);void init_it(int *argc, char ***argv) { mpi_err = MPI_Init(argc,argv); mpi_err = MPI_Comm_size( MPI_COMM_WORLD, &numnodes ); mpi_err = MPI_Comm_rank(MPI_COMM_WORLD, &myid);}
37San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
int main(int argc,char *argv[]){ int *sray,*rray; int *scounts,*rcounts; int ssize,rsize,i,k,j; float z; init_it(&argc,&argv); scounts=(int*)malloc(sizeof(int)*numnodes); rcounts=(int*)malloc(sizeof(int)*numnodes); /*! seed the random number generator with a! different number on each processor*/ seed_random(myid);/* find data to send */ for(i=0;i<numnodes;i++){ random_number(&z); scounts[i]=(int)(10.0*z)+1; } printf("myid= %d scounts=",myid); for(i=0;i<numnodes;i++)
printf("%d ",scounts[i]);
printf("\n");
38San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
/* send the data */ mpi_err = MPI_Alltoall( scounts,1,MPI_INT, rcounts,1,MPI_INT, MPI_COMM_WORLD); printf("myid= %d rcounts=",myid); for(i=0;i<numnodes;i++) printf("%d ",rcounts[i]); printf("\n"); mpi_err = MPI_Finalize();}
void seed_random(int id){ srand((unsigned int)id);}
void random_number(float *z){ int i; i=rand(); *z=(float)i/32767;
}
39San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Output of previous fortran code on 4 procs
ultra:/work/majumdar/examples/mpi % bsub -q hpc -m ultra -I -n 4 a.outJob <48059> is submitted to queue <hpc>.<<Waiting for dispatch ...>><<Starting on ultra>> myid= 1 scounts= 5 2 4 4 myid= 1 rcounts= 1 2 2 3 myid= 2 scounts= 9 2 2 6 myid= 2 rcounts= 7 4 2 11 myid= 3 scounts= 3 3 11 8 myid= 3 rcounts= 2 4 6 8 myid= 0 scounts= 11 1 7 2 myid= 0 rcounts= 11 5 9 3
--------------------------------------------11 1 7 2 11 5 9 35 2 4 4 1 2 2 39 2 2 6 7 4 2 113 3 11 8 2 4 6 8
40San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
The variable or “V” operators
• A collection of very powerful but difficult to setup global communication routines
• MPI_Gatherv: Gather different amounts of data from each processor to the root processor
• MPI_Alltoallv: Send and receive different amounts of data form all processors
• MPI_Allgatherv: Gather different amounts of data from each processor and send all data to each
• MPI_Scatterv: Send different amounts of data to each processor from the root processor
• We discuss MPI_Gatherv and MPI_Alltoallv
41San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI_Gatherv• C
– int MPI_Gatherv (&sendbuf, sendcnts, sendtype, &recvbuf, &recvcnts, &rdispls,recvtype, comm);
• Fortran– MPI_Gatherv (sendbuf, sendcnts, sendtype, recvbuf,
recvcnts, rdispls, recvtype, comm, ierror)
• Parameters:– Recvcnts is now an array
– Rdispls is a displacement
· See attached codes
42San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI GatherV
rank 0 = root rank 1 rank 2
1 2 3
sendbuf 2 3
sendbuf 3
sendbuf
Recvcounts(0) 1 0 = displs(0)
Recvcounts(1) 2 1 = displs(1)
2 2
Recvcounts(2) 3 3 = displs(2)
3 4
3 5
Recvbuf
43San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI Gatherv codeSample program:
include ‘mpif.h’integer isend(3), irecv(3)integer ircnt(0:2), idisp(0:2)data icrnt/1,2,3/ idisp/0,1,3/call mpi_init(ierr)call mpi_comm_size(MPI_COMM_WORLD, nprocs,ierr)call mpi_comm_rank(MPI_COMM_WORLD,myrank,ierr)do I = 1,myrank+1
isend(I) = myrank+1enddoiscnt = myrank + 1call MPI_GATHERV(isend,iscnt,MPI_INTEGER,irecv,ircnt,idisp,MPI_INTEGER
& 0,MPI_COMM_WORLD, ierr)if (myrank .eq. 0) then
print *, ‘irecv =‘, irecvendifcall MPI_FINALIZE(ierr)end
Sample execution:% bsub –q hpc –m ultra –I –n 3 ./a.out% 0: irecv = 1 2 2 3 3 3
44San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI_Alltoallv
• Send and receive different amounts of data form all processors
• C– int MPI_Alltoallv (&sendbuf, &sendcnts, &sdispls,
sendtype, &recvbuf, &recvcnts, &rdispls, recvtype, comm );
• Fortran– Call MPI_Alltoallv(sendbuf, sendcnts, sdispls,
sendtype, recvbuf, recvcnts, rdispls,recvtype, comm,ierror);
• See attached code
45San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI ALLTOALLV
rank0 rank1 rank2
Sendcounts(0) 1 4 7 0=sdispls(0)
Sendcounts(1) 2 5 8 1=sdispls(1)
2 5 8 2
Sendcounts(3) 3 6 9 3=sdispls(2)
3 6 9 4
3 6 9 5
sendbuf sendbuf
Recvcounts(0) 1 2 3 0=rdispls(0)
Recvcounts(1) 4 2 3 1
Recvcounts(3) 7 5 3 2
recvbuf 5 6 3=rdispls(1)
8 6 4
8 6 5
recvbuf 9 6=rdispls(2)
9 7
9 8
46San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI ALLTOALLV
proc
ircnt 0 1 2
0 1 2 3
1 1 2 3
2 1 2 3
proc
irdsp 0 1 2
0 0 0 0
1 1 2 3
2 2 4 6
47San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI ALLTOALLVProgram alltoallv
include ‘mpif.h’
integer isend(6), irecv(9)
integer iscnt(0:2), isdsp(0:2), ircnt(0), irdsp(0:2)
data isend/1,2,2,3,3,3/
data iscnt/1,2,3/ isdsp/0,1,3/
call MPI_INIT(ierr)
call MPI_COMM_SIZE(MPI_COMM_WORLD,nprocs, ierr)call MPI_COMM_RANK(MIP_COMM_WORLD,myrank, ierr)do i = 1,6
isend(i) = isend(i) + nprocs*myrankenddo
do i = 0, nprocs – 1ircnt(i) = myrank + 1irdsp(i) = i* (myrank + 1)
enddo
print*, ‘isend=‘, isendcall MP_FLUSH(1)call MPI_ALLTOALLV(isend,iscnt,isdsp,MPI_INTEGER,irecv, ircnt, irdsp,MPI_INTEGER,
MPI_COMM_WORLD, ierr)print*, ‘irecv=‘,irecvcall MPI_FINALIZE(ierr)end
48San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI ALLTOALLV
Sample execution of mpialltoallv program:
% bsub –q hpc –m ultra –I –n 3
% 0: isend = 1 2 2 3 3 3
1: isend = 4 5 5 6 6 6
2: isend = 7 8 8 9 9 9
0: irecv = 1 4 7 0 0 0 0 0 0
1: irecv = 2 2 5 5 8 8 0 0 0
2: irecv = 3 3 3 6 6 6 9 9 9
49San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Derived types
• C and Fortran 90 have the ability to define arbitrary data types that encapsulate reals, integers, and characters.
• MPI allows you to define message data types corresponding to your data types
• Can use these data types just as default types
50San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Derived types, Three main classifications:
• Contiguous Vectors: enable you to send contiguous blocks of the same type of data lumped together
• Noncontiguous Vectors: enable you to send noncontiguous blocks of the same type of data lumped together
• Abstract types: enable you to (carefully) send C or Fortran 90 structures, don't send pointers
51San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Derived types, how to use them
• Three step process– Define the type using
• MPI_TYPE_CONTIGUOUS for contiguous vectors
• MPI_TYPE_VECTOR for noncontiguous vectors
• MPI_TYPE_STRUCT for structures
– Commit the type using• MPI_TYPE_COMMIT
– Use in normal communication calls• MPI_Send(buffer, count, MY_TYPE,
destination,tag, MPI_COMM_WORLD, ierr)
52San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI_TYPE_CONTIGUOUS• Defines a new data type of length count elements
from your old data type• C
– MPI_TYPE_CONTIGUOUS(int count, old_type, &new_type)
• Fortran– Call MPI_TYPE_CONTIGUOUS(count, old_type,
new_type, ierror)
• Parameters– Old_type: your base type
– New_type: a type count elements of Old_type
• See attached codes
53San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI_TYPE_CONTIGUOUSSample program:
program type_contiguousinclude ‘mpif.h’integer ibuf(20)call MPI_INIT(ierr)call MPI_COMM_SIZE(MPI_COMM_WORLD,nprocs,ierr)call MPI_COMM_RANK(MPI_COMM_WORLD,myrank,ierr)if (myrank .eq. 0) then do i = 1,20
ibuf(i) = I enddoendifcall MPI_TYPE_CONTIGUOUS(3,MPI_INTEGER,inewtype, ierr)call MPI_TYPE_COMMIT(inewtype, ierr)call MPI_BCAST(ibuf,3,inewtype,0,MPI_COMM_WORLD, ierr)print*, ‘ibuf=‘,ibufcall MPI_FINALIZE(ierr)end
Sample execution:% bsub –q hpc –m ultra –I in 2 a.out% 0 : ibuf =1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 1: ibuf = 1 2 3 4 5 6 7 8 9 0 0 0 0 0 0 0 0 0 0 0
54San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI_TYPE_VECTOR
• Defines a datatype which consists of count blocks each of length blocklength and stride displacement between blocks
• C– MPI_TYPE_VECTOR(count, blocklength, stride,
old_type, *new_type)
• Fortran– Call MPI_TYPE_VECTOR(count, blocklength, stride,
old_type, new_type, ierror)
• See attached codes
55San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
module mpi
!DEC$ NOFREEFORM
include "mpif.h"
!DEC$ FREEFORM
end module
!Shows how to use MPI_Type_vector to send noncontiguous blocks of data
!and MPI_Get_count and MPI_Get_elements to see the number of elements sent
program do_vect
use mpi
! include "mpif.h"
integer , parameter :: size=24
integer myid, ierr,numprocs
real*8 svect(0:size),rvect(0:size)
integer i,bonk1,bonk2,numx,stride,extent
integer MY_TYPE
integer status(MPI_STATUS_SIZE)
call MPI_INIT( ierr )
call MPI_COMM_RANK( MPI_COMM_WORLD, myid, ierr )
call MPI_COMM_SIZE( MPI_COMM_WORLD, numprocs, ierr )
stride=5
numx=(size+1)/stride
56San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
extent = 1
if(myid == 1)write(*,*)"numx=",numx," extent=",extent," stride=",stride
call MPI_Type_vector(numx,extent,stride,MPI_DOUBLE_PRECISION,MY_TYPE,ier
r)
call MPI_Type_commit(MY_TYPE, ierr )
if(myid == 0)then
do i=0,size
svect(i)=i
enddo
call MPI_Send(svect,1,MY_TYPE,1,100,MPI_COMM_WORLD,ierr)
endif
if(myid == 1)then
do i=0,size
rvect(i)=-1
enddo
call MPI_Recv(rvect,1,MY_TYPE,0,100,MPI_COMM_WORLD,status,ierr)
endif
if(myid == 1)then
call MPI_Get_count(status,MY_TYPE,bonk1, ierr )
call MPI_Get_elements(status,MPI_DOUBLE_PRECISION,bonk2,ierr)
57San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
write(*,*)"got ", bonk1," elements of type MY_TYPE"
write(*,*)"which contained ", bonk2," elements of type MPI_DOUBLE_PREC
ISION"
do i=0,size
if(rvect(i) /= -1)write(*,'(i2,f4.0)')i,rvect(i)
enddo
endif
call MPI_Finalize(ierr )
end program
! output
! numx= 5 extent= 1 stride= 5
! got 1 elements of type MY_TYPE
! which contained 5 elements of type MPI_DOUBLE_PRECISION
! 0 0.
! 5 5.
! 10 10.
! 15 15.
! 20 20.
58San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI_TYPE_STRUCT
• Defines a MPI datatype which maps to a user defined derived datatype
• C– int MPI_TYPE_STRUCT(count, &array_of_blocklengths,
&array_of_displacement, &array_of_types, &newtype);
• Fortran– Call MPI_TYPE_STRUCT(count, array_of_blocklengths,
array_of_displacement, array_of_types, newtype,ierror)
59San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI_TYPE_STRUCT
• Parameters: – [IN count] # of old types in the new type (integer)
– [IN array_of_blocklengths] how many of each type in new structure (integer)
– [IN array_of_types] types in new structure (integer)
– [IN array_of_displacement] offset in bytes for the beginning of each group of types (integer)
– [OUT newtype] new datatype (handle) – Call MPI_TYPE_STRUCT(count, array_of_blocklengths,
array_of_displacement,array_of_types, newtype,ierror)
60San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Derived Data type Example
Consider the data type or structure consisting of 3 MPI_DOUBLE_PRECISION 10 MPI_INTEGER 2 MPI_LOGICAL Creating the MPI data structure matching this C/Fortran structure is a three step process Fill the descriptor arrays: B - blocklengths T - types D - displacements Call MPI_TYPE_STRUCT to create the MPI data structure Commit the new data type using MPI_TYPE_COMMIT
61San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Derived Data type Example
• Consider the data type or structure consisting of– 3 MPI_DOUBLE_PRECISION
– 10 MPI_INTEGER
– 2 MPI_LOGICAL
• To create the MPI data structure matching this C/Fortran structure– Fill the descriptor arrays:
• B - blocklengths
• T - types
• D - displacements
– Call MPI_TYPE_STRUCT
62San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Derived Data type Example (continued)! t contains the types that! make up the structure
t(1)=MPI_DOUBLE_PRECISION t(2)=MPI_INTEGER t(3)=MPI_LOGICAL! b contains the number of each type
b(1)=3;b(2)=10;b(3)=2! d contains the byte offset of! the start of each type
d(1)=0;d(2)=24;d(3)=64 call MPI_TYPE_STRUCT(3,b,d,t, MPI_CHARLES,mpi_err)
MPI_CHARLES is our new data type
63San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI_Type_commit
• Before we use the new data type we call MPI_Type_commit
• C– MPI_Type_commit(MPI_CHARLES)
• Fortran– Call MPI_Type_commit(MPI_CHARLES,ierr)
64San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Communicators
• A communicator is a parameter in all MPI message passing routines
• A communicator is a collection of processors that can engage in communication
• MPI_COMM_WORLD is the default communicator that consists of all processors
• MPI allows you to create subsets of communicators
65San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
Why Communicators?
• Isolate communication to a small number of processors
• Useful for creating libraries
• Different processors can work on different parts of the problem
• Useful for communicating with "nearest neighbors"
66San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI_Comm_split
• Provides a short cut method to create a collection of communicators
• All processors with the "same color" will be in the same communicator
• Index gives rank in new communicator
• Fortran– call MPI_COMM_SPLIT(OLD_COMM, color, index,
NEW_COMM, mpi_err)
• C– MPI_Comm_split(OLD_COMM, color, index, &NEW_COMM)
67San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI_Comm_split• Split odd and even processors into 2 communicators
Program comm_splitinclude "mpif.h"Integer color,zero_onecall MPI_INIT( mpi_err )call MPI_COMM_SIZE( MPI_COMM_WORLD, numnodes, mpi_err )call MPI_COMM_RANK( MPI_COMM_WORLD, myid, mpi_err )color=mod(myid,2) !color is either 1 or 0call MPI_COMM_SPLIT(MPI_COMM_WORLD,color,myid,NEW_COMM,mpi_err)call MPI_COMM_RANK( NEW_COMM, new_id, mpi_err )call MPI_COMM_SIZE( NEW_COMM, new_nodes, mpi_err )Zero_one = -1If(new_id==0)Zero_one = colorCall MPI_Bcast(Zero_one,1,MPI_INTEGER,0, NEW_COMM,mpi_err)If(zero_one==0)write(*,*)"part of even processor communicator"If(zero_one==1)write(*,*)"part of odd processor communicator"Write(*,*)"old_id=", myid, "new_id=", new_idCall MPI_FINALIZE(mpi_error)End program
68San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
MPI_Comm_split
• Split odd and even processors into 2 communicators0: part of even processor communicator0: old_id= 0 new_id= 0
2: part of even processor communicator2: old_id= 2 new_id= 1
1: part of odd processor communicator1: old_id= 1 new_id= 0
3: part of odd processor communicator3: old_id= 3 new_id= 1
69San DIEGO SUPERCOMPUTER CENTER
NATIONAL PARTNERSHIP FOR ADVANCED COMPUTATIONAL INFRASTRUCTURE
#include "mpi.h"#include <math.h>int main(argc,argv)int argc;char *argv[];{ int myid, numprocs; int color,Zero_one,new_id,new_nodes; MPI_Comm NEW_COMM; MPI_Init(&argc,&argv); MPI_Comm_size(MPI_COMM_WORLD,&numprocs); MPI_Comm_rank(MPI_COMM_WORLD,&myid); color=myid % 2; MPI_Comm_split(MPI_COMM_WORLD,color,myid,&NEW_COMM); MPI_Comm_rank( NEW_COMM, &new_id); MPI_Comm_size( NEW_COMM, &new_nodes); Zero_one = -1; if(new_id==0)Zero_one = color; MPI_Bcast(&Zero_one,1,MPI_INT,0, NEW_COMM); if(Zero_one==0)printf("part of even processor communicator \n"); if(Zero_one==1)printf("part of odd processor communicator \n"); printf("old_id= %d new_id= %d\n", myid, new_id); MPI_Finalize();}