⭐ 欢迎来到虫虫下载站! | 📦 资源下载 📁 资源专辑 ℹ️ 关于我们
⭐ 虫虫下载站

📄 jgfbcastbench.java

📁 MPI for java for Distributed Programming
💻 JAVA
字号:
/***************************************************************************                                                                         **         Java Grande Forum Benchmark Suite - MPJ Version 1.0             **                                                                         **                            produced by                                  **                                                                         **                  Java Grande Benchmarking Project                       **                                                                         **                                at                                       **                                                                         **                Edinburgh Parallel Computing Centre                      **                                                                         * *                email: epcc-javagrande@epcc.ed.ac.uk                     **                                                                         **                                                                         **      This version copyright (c) The University of Edinburgh, 2001.      **                         All rights reserved.                            **                                                                         ***************************************************************************/package jgf_mpj_benchmarks.section1;import java.io.*;import jgf_mpj_benchmarks.jgfutil.*; //import jgfutil.*; import mpi.*;public class JGFBcastBench implements JGFSection1{  public  static  int nprocess;  public  static  int rank;  private static final int INITSIZE = 1;  private static final int MAXSIZE =  1000000;  private static final double TARGETTIME = 10.0;  private static final int MLOOPSIZE = 25;  private static final int SMAX = 5000000;  private static final int SMIN = 4;  public JGFBcastBench(int rank, int nprocess) {	this.rank = rank;          this.nprocess=nprocess;  }  public void JGFrun() throws MPIException {    int size,i,l,m_size;    double logsize;    double b_time;     b_time = 0.0;    double [] time = new double[1];    m_size = 0;    logsize = Math.log((double) SMAX) - Math.log((double) SMIN);/* Bcast an array of doubles *//* Create the timers */       if(rank==0){        JGFInstrumentor.addTimer("Section1:Bcast:Double", "bytes");        JGFInstrumentor.addTimer("Section1:Bcast:Barrier", "barriers");      }/* loop over no of different message sizes */    for(l=0;l<MLOOPSIZE;l++){/* Initialize the sending data */      m_size = (int)(Math.exp(Math.log((double)SMIN)+(double) ((double) l/(double) MLOOPSIZE*logsize)));      double [] arr = new double[m_size];      time[0] = 0.0;      size=INITSIZE;      MPI.COMM_WORLD.Barrier();/* Start the timer */      while (time[0] < TARGETTIME && size < MAXSIZE){        if(rank==0){          JGFInstrumentor.resetTimer("Section1:Bcast:Double");          JGFInstrumentor.startTimer("Section1:Bcast:Double");        }/* Carryout the broadcast operation */        for (int k=0; k<size; k++){          MPI.COMM_WORLD.Bcast(arr,0,arr.length,MPI.DOUBLE,0);          MPI.COMM_WORLD.Barrier();        }/* Stop the timer */        if(rank==0){          JGFInstrumentor.stopTimer("Section1:Bcast:Double");           time[0] = JGFInstrumentor.readTimer("Section1:Bcast:Double");           JGFInstrumentor.addOpsToTimer("Section1:Bcast:Double",(double) size*arr.length*8);         }/* Broadcast time to the other processes  */        MPI.COMM_WORLD.Barrier();        MPI.COMM_WORLD.Bcast(time,0,1,MPI.DOUBLE,0);        size *=2;      }         size /=2;/* determine the cost of the Barrier, subtract the cost and write out the performance time */      MPI.COMM_WORLD.Barrier();       if(rank==0) {        JGFInstrumentor.resetTimer("Section1:Bcast:Barrier");        JGFInstrumentor.startTimer("Section1:Bcast:Barrier");      }      for (int k=0; k<size; k++){        MPI.COMM_WORLD.Barrier();      }      if(rank==0) {        JGFInstrumentor.stopTimer("Section1:Bcast:Barrier");        b_time = JGFInstrumentor.readTimer("Section1:Bcast:Barrier");        JGFInstrumentor.addTimeToTimer("Section1:Bcast:Double", -b_time);        JGFInstrumentor.printperfTimer("Section1:Bcast:Double",arr.length);       }    }/* Bcast an array of objects containing a double *//* Create the timer */    if(rank==0){      JGFInstrumentor.addTimer("Section1:Bcast:Object", "objects");    }/* loop over no of different message sizes */    for(l=0;l<MLOOPSIZE;l++){/* Initialize the sending data */      m_size = (int)(Math.exp(Math.log((double)SMIN)+(double) ((double) l/(double) MLOOPSIZE*logsize)));      obj_double [] arr_obj = new obj_double[m_size];      for(int k=0;k<m_size;k++){       arr_obj[k] = new obj_double(0.0);      }      time[0] = 0.0;      size=INITSIZE;      MPI.COMM_WORLD.Barrier();/* Start the timer */      while (time[0] < TARGETTIME && size < MAXSIZE){        if(rank==0){          JGFInstrumentor.resetTimer("Section1:Bcast:Object");          JGFInstrumentor.startTimer("Section1:Bcast:Object");        }/* Carryout the broadcast operation */        for (int k=0; k<size; k++){          MPI.COMM_WORLD.Bcast(arr_obj,0,arr_obj.length,MPI.OBJECT,0);          MPI.COMM_WORLD.Barrier();        }/* Stop the timer */        if(rank==0){          JGFInstrumentor.stopTimer("Section1:Bcast:Object");          time[0] = JGFInstrumentor.readTimer("Section1:Bcast:Object");          JGFInstrumentor.addOpsToTimer("Section1:Bcast:Object",(double) size*arr_obj.length);        }/* Broadcast time to the other processes */        MPI.COMM_WORLD.Barrier();        MPI.COMM_WORLD.Bcast(time,0,1,MPI.DOUBLE,0);        size *=2;      }        size /=2;/* determine the cost of the Barrier, subtract the cost and write out the performance time */      MPI.COMM_WORLD.Barrier();      if(rank==0) {        JGFInstrumentor.resetTimer("Section1:Bcast:Barrier");        JGFInstrumentor.startTimer("Section1:Bcast:Barrier");      }      for (int k=0; k<size; k++){        MPI.COMM_WORLD.Barrier();      }      if(rank==0) {        JGFInstrumentor.stopTimer("Section1:Bcast:Barrier");        b_time = JGFInstrumentor.readTimer("Section1:Bcast:Barrier");        JGFInstrumentor.addTimeToTimer("Section1:Bcast:Object", -b_time);        JGFInstrumentor.printperfTimer("Section1:Bcast:Object",arr_obj.length);      }    }  }  public static void main(String[] argv) throws MPIException{  }  public JGFBcastBench(String[] argv) throws Exception {/* Initialise MPI */     MPI.Init(argv);     rank = MPI.COMM_WORLD.Rank();     nprocess = MPI.COMM_WORLD.Size();     if(rank==0){     JGFInstrumentor.printHeader(1,0,nprocess);     }     JGFBcastBench bc = new JGFBcastBench(rank, nprocess);     bc.JGFrun();/* Finalise MPI */     MPI.Finalize();  }}

⌨️ 快捷键说明

复制代码 Ctrl + C
搜索代码 Ctrl + F
全屏模式 F11
切换主题 Ctrl + Shift + D
显示快捷键 ?
增大字号 Ctrl + =
减小字号 Ctrl + -