mpich/2008-07-20_YM_MPI_Course – Grid Architecture

wiki:mpich/2008-07-20_YM_MPI_Course

Context Navigation

使用你的帳號登入: ym

login as: ym000
ym000@140.129.162.12's password: ******
ym000@bio001:~$ ssh-keygen -t rsa

產生 SSH 認證金鑰

Generating public/private rsa key pair.
Enter file in which to save the key (/home/ym000/.ssh/id_rsa):
Created directory '/home/ym000/.ssh'.
Enter passphrase (empty for no passphrase):
Enter same passphrase again:
Your identification has been saved in /home/ym000/.ssh/id_rsa.
Your public key has been saved in /home/ym000/.ssh/id_rsa.pub.
The key fingerprint is:
2a:6c:05:f8:24:38:db:79:b9:4f:0c:74:da:c5:16:05 ym000@bio001

進行金鑰交換
```
ym000@bio001:~$ cp .ssh/id_rsa.pub .ssh/authorized_keys
```
- [備註] 因為是 DRBL 環境, 因此每一台都已經有 .ssh/authorized_keys, 如果不是 DRBL 環境, 你必須要自己手動把 .ssh/id_rsa.pub 拷貝到每一台 Compute Node 的 .ssh/authorized_keys
```
ym000@bio001:~$ for ((i=2;i<=7;i++)); do scp .ssh/id_rsa.pub ym000@192.168.192.$i:.ssh/authorized_keys ; done
```

設定 MPD 設定檔跟 MPI 的執行檔路徑

ym000@bio001:~$ echo "MPD_SECRETWORD=${user}$$" > ~/.mpd.conf
ym000@bio001:~$ chmod 600 .mpd.conf
ym000@bio001:~$ for ((i=2;i<=7;i++)); do echo "192.168.129.$i" >> mpd.hosts; done
ym000@bio001:~$ export PATH=$PATH:/opt/mpich2/bin
ym000@bio001:~$ which mpdboot
/opt/mpich2/bin/mpdboot

設定 dsh (distributed shell)*, 我們可以使用 dsh 指令逐台執行.

ym000@bio001:~$ mkdir -p .dsh/
ym000@bio001:~$ cp mpd.hosts .dsh/machines.list
ym000@bio001:~$ dsh -a hostname
bio001002
bio001003
The authenticity of host '192.168.129.4 (192.168.129.4)' can't be established.
RSA key fingerprint is f0:4b:6f:52:3c:0b:f4:8b:1c:a0:33:4a:e2:15:e0:5a.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added '192.168.129.4' (RSA) to the list of known hosts.
bio001004
The authenticity of host '192.168.129.5 (192.168.129.5)' can't be established.
RSA key fingerprint is 09:9b:25:5e:9c:a8:9a:dd:35:ee:f0:54:6a:11:b7:90.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added '192.168.129.5' (RSA) to the list of known hosts.
bio001005
The authenticity of host '192.168.129.6 (192.168.129.6)' can't be established.
RSA key fingerprint is 40:10:50:38:2c:f0:0b:f7:11:85:a3:41:d9:fb:ac:7d.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added '192.168.129.6' (RSA) to the list of known hosts.
bio001006
The authenticity of host '192.168.129.7 (192.168.129.7)' can't be established.
RSA key fingerprint is 07:95:b7:f8:a5:9c:c7:21:84:d0:5b:f4:5f:db:0b:a6.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added '192.168.129.7' (RSA) to the list of known hosts.
bio001007
ym000@bio001:~$ dsh -a hostname
bio001002
bio001003
bio001004
bio001005
bio001006
bio001007

用使用者的身分執行 mpd
```
ym000@bio001:~$ mpdboot -n 7
```

用 mpdtrace 檢查 mpd 執行狀態

ym000@bio001:~$ mpdtrace 
bio001
bio001005
bio001004
bio001003
bio001002
bio001007
bio001006

用 mpdringtest 做 mpd 訊息傳遞效能測試

ym000@bio001:~$ mpdringtest 1000
time for 1000 loops = 0.648007154465 seconds

用 mpiexec 執行 cpi 範例程式

ym000@bio001:~$ mpiexec -n 3 /opt/mpich2/share/mpich2/examples/cpi
Process 0 of 1 is on bio001
pi is approximately 3.1415926544231341, Error is 0.0000000008333410
wall clock time = 0.000284
Process 0 of 1 is on bio001
pi is approximately 3.1415926544231341, Error is 0.0000000008333410
wall clock time = 0.000295
Process 0 of 1 is on bio001
pi is approximately 3.1415926544231341, Error is 0.0000000008333410
wall clock time = 0.000294

貼上 test1.c

ym000@bio001:~$ cat << EOF > test1.c
> #include <stdio.h>
> #include <mpi.h>
> main (int argc, char **argv)
> {
>   int rank, size, len;
>   char name[MPI_MAX_PROCESSOR_NAME];
>   MPI_Init(&argc, &argv);
>   int myid, numprocs;
>
>   /* 取得 node 總數 */
>   MPI_Comm_size(MPI_COMM_WORLD,&numprocs);
>   /* 取得本身 node id / rank */
>   MPI_Comm_rank(MPI_COMM_WORLD,&myid);
  MPI_Get_processor_name(name, &len);
>   /* 取得本身 host name */
>   MPI_Get_processor_name(name, &len);
>   printf("This is machine %d of %d name = %s\n", myid, numprocs, name);
>
>   MPI_Finalize();
> }
> EOF

用 mpicc 編譯 test1.c
```
ym000@bio001:~$ mpicc -o test1 test1.c
```

用 mpiexec 執行 test1 程式

ym000@bio001:~$ mpiexec -n 1 ./test1
This is machine 0 of 1 name = bio001
ym000@bio001:~$ mpiexec -n 12 ./test1
This is machine 0 of 12 name = bio001
This is machine 1 of 12 name = bio001004
This is machine 2 of 12 name = bio001005
This is machine 3 of 12 name = bio001003
This is machine 4 of 12 name = bio001002
This is machine 5 of 12 name = bio001007
This is machine 6 of 12 name = bio001006
This is machine 7 of 12 name = bio001
This is machine 8 of 12 name = bio001004
This is machine 11 of 12 name = bio001002
This is machine 9 of 12 name = bio001005
This is machine 10 of 12 name = bio001003

here is test1.c

#include <stdio.h>
#include <mpi.h>
main (int argc, char **argv)
{
  int rank, size, len;
  char name[MPI_MAX_PROCESSOR_NAME];
  MPI_Init(&argc, &argv);
  int myid, numprocs;

  /* 取得 node 總數 */
  MPI_Comm_size(MPI_COMM_WORLD,&numprocs);
  /* 取得本身 node id / rank */
  MPI_Comm_rank(MPI_COMM_WORLD,&myid);
  /* 取得本身 host name */
  MPI_Get_processor_name(name, &len);
  printf("This is machine %d of %d name = %s\n", myid, numprocs, name);

  MPI_Finalize();
}

here is test2.c

#include <mpi.h>
#include <stdio.h>
main(int argc,char **argv) { 
  int n, myrank, numprocs;
  MPI_Status status;
  MPI_Init(&argc,&argv);
  MPI_Comm_size(MPI_COMM_WORLD,&numprocs);
  MPI_Comm_rank(MPI_COMM_WORLD,&myrank);

  /* node 0 will send the first message */
  if(myrank == 0) {
    n = myrank;
    MPI_Send(&n, 1, MPI_INT, 1, 99, MPI_COMM_WORLD);
    printf("[Ndde %d]「%d」 >> [Node %d]\n\n", myrank, n, myrank+1);
  }

  /* node 1 to node n-2 will send message to the next node */
  if(myrank>0 && myrank<numprocs-1) {
    MPI_Recv(&n, 1, MPI_INT, myrank-1, 99, MPI_COMM_WORLD, &status);
    printf("[Node %d] << 「%d」[Node %d]\n", myrank, n, status.MPI_SOURCE);
    n = myrank; MPI_Send(&n, 1, MPI_INT, myrank+1, 99, MPI_COMM_WORLD);
    printf("[Ndde %d]「%d」 >> [Node %d]\n\n", myrank, n, myrank+1);
  }

 /* the final node n-1 will not send any message but receive*/
  if(myrank==numprocs-1) {
    MPI_Recv(&n, 1, MPI_INT, myrank-1, 99, MPI_COMM_WORLD, &status);
    printf("[Node %d] << 「%d」[Node %d]\n", myrank, n, status.MPI_SOURCE);
    }

  MPI_Finalize();
}

here is test3.c

/* Program:
 *   每個 node 將訊息傳送給 node 0，由，node 0 統一印出
 * History:
 *   2008-06-12 BETA
 *   2008-06-17 更改顯示方式，並增加註解
 */

#include <stdio.h>
#include <mpi.h>
#include <string.h>

main(int argc, char **argv)
{
  int myrank, i, numprocs;
  char message[20];
  MPI_Status status;
  MPI_Init(&argc, &argv);
  MPI_Comm_size(MPI_COMM_WORLD, &numprocs);
  MPI_Comm_rank(MPI_COMM_WORLD, &myrank);

  /* Node 0 will do the following */
  if(myrank == 0)
  {
    /* receive messages from other nodes */
    for(i = 1; i < numprocs; i++)
    {
      MPI_Recv(message, 20, MPI_CHAR, i, 99, MPI_COMM_WORLD, &status);
      printf("[Node 0] << 「%s」[Node %d] \n", message, status.MPI_SOURCE);
    }
  }

  /* other Nodes will do the following */
  if(myrank != 0)
  {
    /* send node's rank to Node 0 */
    sprintf(message, "[%d]", myrank);
    MPI_Send(message, 20, MPI_CHAR, 0, 99, MPI_COMM_WORLD);
    printf("[Node %d]「%s」 >> [Node 0]\n", myrank, message);
  }
  MPI_Finalize();
}

here is test4

/* Program:
 *   mpich_example 內建範例，計算 pi 。
 * History:
 *   2008-04-11 BETA
 *   2008-06-19 增加可重複輸入欲計算之精準度
 *   2008-06-23 加入 MPI_Barrier 以確保每個 node 在接受 n 後才執行
 */

#include "mpi.h"
#include <stdio.h>
#include <math.h>
#include <time.h>

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=0;
    double PI25DT = 3.141592653589793238462643;
    double mypi, pi, h, sum, x;
    double startwtime = 0.0, endwtime;
    int  namelen;
    char processor_name[MPI_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 %s\n",
            myid, processor_name);
    n = 0;
    while (!done)
    {
        /* 由 node 0 將使用者輸入的值送給其它的 node  */
        if (myid == 0)
        {
            printf("Enter the number of intervals: (0 quits) ");
            scanf("%d", &n);
            startwtime = MPI_Wtime();
        }

        /* 這非常重要，所有的 node 必需在此同步，才可以收到使用者輸入的 n */
        MPI_Barrier(MPI_COMM_WORLD);

        /* 將 n 送給其它的 node  */
        MPI_Bcast(&n, 1, MPI_INT, 0, MPI_COMM_WORLD);
        if (n == 0)
            done = 1;
        else
        {

            /* 此為計算 pi 的演算法 */
            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;

            /* 將算完的結果傳給 node 0 加總 */
            MPI_Reduce(&mypi, &pi, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_WORLD);
            if (myid == 0)
            {
                printf("pi is approximately %.16f, Error is %.16f\n",
                       pi, fabs(pi - PI25DT));
                endwtime = MPI_Wtime();
                printf("wall clock time = %f\n",
                       endwtime-startwtime);
            }
        }
    }
    MPI_Finalize();
    return 0;
}

Last modified 15 years ago Last modified on Jul 5, 2009, 2:02:43 PM

Download in other formats:

Plain Text