基于消息傳遞的并行計算(MPICHG)

1、MPICH-G2Grid-Enabled Implementation of MPI2The Features of MPICH-G2nMPICH-G2 hides heterogeneity by using Globus Toolkit services for such purposes as: qauthentication, authorization, executable staging, process creation, process monitoring, process control, communication, redirection of standard in

2、put and output and remote file accessnMPICHG enables the use of several different MPI features for user management of heterogeneityqMPIs asynchronous operations hide latency qMPIs communicator construct represent the hierarchical structure of heterogeneous systems3M MP PI IC CH H A Ar rc ch hi it te

3、 ec ct tu ur re eT Th he e M MP PI I i in nt te er rf fa ac ce e (defined by the MPI standards)T Th he e M MP PI IC CH H l la ay ye er r(implements the MPI interface)U Us se er rA Ab bs st tr ra ac ct t D De ev vi ic ce e I In nt te er rf fa ac ce e ( (A AD DI I) )M MP PP PS SM MP PC Cl lu us st te

4、er rA A P Pa ar rt ti ic cu ul la ar r P Pl la at tf fo or rm m4How MPICH-G2 Works5How MPICH-G2 WorksnStep 1 :The user employs the Grid Security Infrastructure (GSI) to obtain a proxy credential that is used to authenticate the user to each site.qThe user may also use the Monitoring and Discovery Se

5、rvice (MDS) to select computes .6How MPICH-G2 WorksnStep 2 :Once authenticated ,the user uses the standard mpirun command to request the creation of an MPI computation .The MPICH-G2 implementation of this command uses the Resource Specification Language (RSL) to describe the job.7How MPICH-G2 Worksn

6、Step 3 : Based on the information found in an RSL script , MPICH-G2 calls a co-allocation library distributed with the Globus Toolkit ,the Dynamically-Updated Request Online Coallocator (DUROC) to schedule and start the application across the various computers specified by the user.nStep 4:The DUROC

7、 library itself uses the Grid Resource Allocation and Management (GRAM) API and protocol to start and subsequently manage a set of subcomputations one for each computer. For each subcomputation DUROC generates a GRAM request to a remote GRAM server which authenticates the user performs local authori

8、zation and then interacts with the local scheduler to initiate the computation.8How MPICH-G2 Works9An Example#include mpi.h#include #include double f( double );double f( double a ) return (4.0 / (1.0 + a*a);int main( int argc, char *argv) int done = 0, n, myid, numprocs, i; double mypi, pi, h, sum,

9、x; double startwtime = 0.0, endwtime; int namelen; char processor_nameMPI_MAX_PROCESSOR_NAME; MPI_Init(&argc,&argv); MPI_Comm_size(MPI_COMM_WORLD,&numprocs); MPI_Comm_rank(MPI_COMM_WORLD,&myid); MPI_Get_processor_name(processor_name,&namelen); fprintf(stderr,Process %d on %sn, my

10、id, processor_name); n = 0;10while (!done) if (myid = 0) if (n=0) n=100; else n=0; startwtime = MPI_Wtime(); MPI_Bcast(&n, 1, MPI_INT, 0, MPI_COMM_WORLD); if (n = 0) done = 1; else h = 1.0 / (double) n; sum = 0.0; for (i = myid + 1; i = n; i += numprocs) x = h * (double)i - 0.5); sum += f(x); 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 %.16fn, pi, fabs(pi - PI25DT); endwtime = MPI_Wtime(); printf(