[Introduction] Example 1 : Game of Life

This material is part of the course Practical Introduction to Parallel Programming and meant for educational purposes only.
Background

Conway's game of life describes the evolution of a \(m \times n\) grid in which each cell has an associated value that is either zero or one.
The value of a cell in the next time step depends on the values of the neighboring cells in the current time step.
The neighbors of a cell \((i,j)\) are the cells \((i \pm 1, j \pm 1)\).
The initial state of the grid completely determines how the grid evolves in time.
The value of a cell transitions according to the following rules:
a non-zero cell becomes zero if it has less than two non-zero neighbors;
a non-zero cell remains non zero if it has two or three non-zero neighbors;
a non-zero cell becomes zero if it has more than three non-zero neighbors;
a zero cell becomes non zero if it has exactly three non-zero neighbors.
Dán dòng lệnh này vào terminal của bạn:
Mật khẩu để kết nối là
Câu hỏi 1: Game of Life

#include <stdio.h>
#include <assert.h>
#include <stdarg.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <mpi.h>

/* ---------------------------------------------------------------------
 * global variables, constants and types
 * ------------------------------------------------------------------ */

#define                   INIT_TAG      1
#define                   EXCHANGE_TAG  2

#define                   DEBUG_SOME    1
#define                   DEBUG_ALL     2

static int                debug_level = 0;

typedef unsigned char     byte_t;

struct                    conf
{
  int                       dims[2];
  int                       steps;
};

struct                    grid
{
  int                       origin     [2];
  int                       local_dims [2];
  int                       global_dims[2];
  byte_t*                   cells;
};

struct                    xdata
{
  int                       req_count;
  int                       xchg_count;
  byte_t*                   send_buffer;
  byte_t*                   recv_buffer;
  int*                      recv_index;
  int*                      send_index;
  MPI_Request*              send_reqs;
  MPI_Request*              recv_reqs;
};

struct                    proc_info
{
  int                       my_rank;
  int                       my_coords[2];
  int                       proc_dims[2];
};

typedef struct conf       conf_t;
typedef struct grid       grid_t;
typedef struct xdata      xdata_t;
typedef struct proc_info  proc_info_t;

/* ---------------------------------------------------------------------
 * error()
 * ------------------------------------------------------------------ */

void                      error

  ( const char*             fmt, ... )

{
  va_list  ap;

  va_start  ( ap, fmt );
  fprintf   ( stderr, "\n*** ERROR: " );
  vfprintf  ( stderr, fmt, ap );
  fprintf   ( stderr, "\n\n" );
  va_end    ( ap );
  MPI_Abort ( MPI_COMM_WORLD, 1 );
  exit      ( 1 );
}

/* ---------------------------------------------------------------------
 * allocate()
 * ------------------------------------------------------------------ */

void*                     allocate

  ( size_t                  nmemb,
    int                     count )

{
  void*  ptr = NULL;

  if ( count > 0 )
  {
    ptr = calloc ( nmemb, (size_t) count );

    if ( ! ptr )
    {
      error ( "failed to allocate %ld bytes",
              (long) (count * nmemb) );
    }
  }

  return ptr;
}

/* ---------------------------------------------------------------------
 * kill_grid()
 * ------------------------------------------------------------------ */

void                      kill_grid

  ( grid_t*                 grid )

{
  free   ( grid->cells );
  memset ( grid, 0x0, sizeof(*grid) );
}

/* ---------------------------------------------------------------------
 * kill_xdata()
 * ------------------------------------------------------------------ */

void                      kill_xdata

  ( xdata_t*                xdata )

{
  int  i;

  for ( i = 0; i < xdata->req_count; i++ )
  {
    MPI_Request_free ( &xdata->send_reqs[i] );
    MPI_Request_free ( &xdata->recv_reqs[i] );
  }

  free   ( xdata->recv_index );
  free   ( xdata->send_index );
  free   ( xdata->recv_buffer );
  free   ( xdata->send_buffer );
  free   ( xdata->recv_reqs );
  free   ( xdata->send_reqs );

  memset ( xdata, 0x0, sizeof(*xdata) );
}

/* ---------------------------------------------------------------------
 * proc_rank()
 * ------------------------------------------------------------------ */

int                       proc_rank

  ( int                     ipos,
    int                     jpos,
    const proc_info_t*      info )

{
  const int  m = info->proc_dims[0];
  const int  n = info->proc_dims[1];

  if ( (ipos < 0) || (ipos >= m) ||
       (jpos < 0) || (jpos >= n) )
  {
    return -1;
  }
  else
  {
    return (ipos + jpos * m);
  }
}

/* ---------------------------------------------------------------------
 * print_grid()
 * ------------------------------------------------------------------ */

void                      print_grid

  ( const grid_t*           grid )

{
  const int      m   = grid->local_dims[0];
  const int      n   = grid->local_dims[1];
  const byte_t*  row = NULL;

  int            i, j;

  putchar ( '+' );

  for ( i = 0; i < m; i++ )
  {
    putchar ( '-' );
  }

  putchar ( '+' );
  putchar ( '\n' );

  for ( j = n; j >= 1; j-- )
  {
    row = grid->cells + (j * (m + 2));

    putchar ( '|' );

    for ( i = 1; i <= m; i++ )
    {
      if ( row[i] )
      {
        putchar ( 'X' );
      }
      else
      {
        putchar ( ' ' );
      }
    }

    putchar ( '|' );
    putchar ( '\n' );
  }

  putchar ( '+' );

  for ( i = 0; i < m; i++ )
  {
    putchar ( '-' );
  }

  putchar ( '+' );
  putchar ( '\n' );
  putchar ( '\n' );
}

/* ---------------------------------------------------------------------
 * parse_args()
 * ------------------------------------------------------------------ */

void                      parse_args

  ( conf_t*                 conf,
    int                     argc,
    char**                  argv )

{
  const int  root = 0;

  char       dummy;
  int        ibuf[3];
  int        iarg;

  /* Set default values. */

  conf->dims[0] = 16;
  conf->dims[1] = 16;
  conf->steps   = 16;

  for ( iarg = 1; iarg < argc; iarg++ )
  {
    if ( strcmp( argv[iarg], "--dims" ) == 0 )
    {
      iarg++;

      if ( iarg >= argc )
      {
        error ( "no dimensions specified" );
      }

      if ( sscanf( argv[iarg], " %d %c %d",
                   &conf->dims[0], &dummy, &conf->dims[1] ) != 3 ||
           (conf->dims[0] <= 0) || (conf->dims[1] <= 0) )
      {
        error ( "invalid dimensions specified" );
      }

      continue;
    }

    if ( strcmp( argv[iarg], "--steps" ) == 0 )
    {
      iarg++;

      if ( iarg >= argc )
      {
        error ( "no number of steps specified" );
      }

      if ( sscanf( argv[iarg], " %d", &conf->steps ) != 1 ||
           (conf->steps < 0) )
      {
        error ( "invalid number of steps specified" );
      }

      continue;
    }

    if ( strcmp( argv[iarg], "--debug" ) == 0 )
    {
      iarg++;

      if ( iarg >= argc )
      {
        error ( "no debug level specified" );
      }

      if ( sscanf( argv[iarg], " %d", &debug_level ) != 1 ||
           (debug_level < 0) )
      {
        error ( "invalid debug level specified" );
      }

      continue;
    }
  }

  ibuf[0] = conf->dims[0];
  ibuf[1] = conf->dims[1];
  ibuf[2] = conf->steps;

  MPI_Bcast ( ibuf, 3, MPI_INT, root, MPI_COMM_WORLD );

  printf ( "Global dimensions : %d x %d\n",
           conf->dims[0], conf->dims[1] );
  printf ( "Number of steps   : %d\n",
           conf->steps);

  if ( debug_level )
  {
    printf ( "Debug level       : %d\n", debug_level );
  }

  printf ( "\n" );
}

/* ---------------------------------------------------------------------
 * recv_conf()
 * ------------------------------------------------------------------ */

void                      recv_conf

  ( conf_t*                 conf )

{
  const int  root = 0;

  int        ibuf[3];

  MPI_Bcast ( ibuf, 3, MPI_INT, root, MPI_COMM_WORLD );

  conf->dims[0] = ibuf[0];
  conf->dims[1] = ibuf[1];
  conf->steps   = ibuf[2];
}

/* ---------------------------------------------------------------------
 * init_grid()
 * ------------------------------------------------------------------ */

void                      init_grid

  ( grid_t*                 grid,
    const conf_t*           conf )

{
  const int  m   = conf->dims[0];
  const int  n   = conf->dims[1];
  byte_t*    row = NULL;

  int        i, j;

  if ( debug_level )
  {
    printf ( "Initializing grid ...\n" );
  }

  grid->origin     [0] = 0;
  grid->origin     [1] = 0;
  grid->local_dims [0] = m;
  grid->local_dims [1] = n;
  grid->global_dims[0] = m;
  grid->global_dims[1] = n;

  /* Note that the grid is bordered by a strip of halo cells. */

  grid->cells = (byte_t*) allocate ( sizeof(*grid->cells), (m + 2) * (n + 2) );

  for ( j = 1; j <= n; j++ )
  {
    row = grid->cells + (j * (m + 2));

    for ( i = 1; i <= m; i++ )
    {
      int  v = rand ();

      if ( v > (RAND_MAX / 2) )
      {
        row[i] = 1;
      }
      else
      {
        row[i] = 0;
      }
    }
  }

  if ( debug_level >= DEBUG_ALL )
  {
    printf     ( "Initial grid:\n\n" );
    print_grid ( grid );
  }
}

/* ---------------------------------------------------------------------
 * init_xdata()
 * ------------------------------------------------------------------ */

void                      init_xdata

  ( xdata_t*                xdata,
    const grid_t*           grid,
    const proc_info_t*      info )

{
  const int     m           = grid->local_dims[0];
  const int     n           = grid->local_dims[1];

  MPI_Request*  send_reqs   = NULL;
  MPI_Request*  recv_reqs   = NULL;

  byte_t*       send_buffer = NULL;
  byte_t*       recv_buffer = NULL;
  int*          send_index  = NULL;
  int*          recv_index  = NULL;

  int           xchg_count  = 2 * m + 2 * n + 4;
  int           req_count   = 8;
  int           iproc       = 0;

  int           xidx[9];
  int           rank[8];
  int           i, j;

  send_reqs   = (MPI_Request*) allocate ( sizeof(*send_reqs),   req_count );
  recv_reqs   = (MPI_Request*) allocate ( sizeof(*recv_reqs),   req_count );
  send_buffer = (byte_t*)      allocate ( sizeof(*send_buffer), xchg_count );
  recv_buffer = (byte_t*)      allocate ( sizeof(*recv_buffer), xchg_count );
  send_index  = (int*)         allocate ( sizeof(*send_index),  xchg_count );
  recv_index  = (int*)         allocate ( sizeof(*recv_index),  xchg_count );

  xchg_count  = 0;

  /* Lower-left cell. */

  rank[iproc]   = proc_rank ( info->my_coords[0] - 1,
                              info->my_coords[1] - 1, info );
  xidx[iproc++] = xchg_count;

  recv_index[xchg_count]   = 0;
  send_index[xchg_count++] = m + 3;

  /* Lower border. */

  rank[iproc]   = proc_rank ( info->my_coords[0],
                              info->my_coords[1] - 1, info );
  xidx[iproc++] = xchg_count;

  for ( i = 1; i <= m; i++ )
  {
    recv_index[xchg_count]   = i;
    send_index[xchg_count++] = i + m + 2;
  }

  /* Lower-right cell. */

  rank[iproc]   = proc_rank ( info->my_coords[0] + 1,
                              info->my_coords[1] - 1, info );
  xidx[iproc++] = xchg_count;

  recv_index[xchg_count]   = m + 1;
  send_index[xchg_count++] = 2 * m + 2;

  /* Right border. */

  rank[iproc]   = proc_rank ( info->my_coords[0] + 1,
                              info->my_coords[1],     info );
  xidx[iproc++] = xchg_count;

  for ( j = 1; j <= n; j++ )
  {
    recv_index[xchg_count]   = m + 1 + j * (m + 2);
    send_index[xchg_count++] = m     + j * (m + 2);
  }

  /* Upper-right cell. */

  rank[iproc]   = proc_rank ( info->my_coords[0] + 1,
                              info->my_coords[1] + 1, info );
  xidx[iproc++] = xchg_count;

  recv_index[xchg_count]   = (m + 2) * (n + 2) - 1;
  send_index[xchg_count++] = (m + 2) * (n + 1) - 2;

  /* Upper border. */

  rank[iproc]   = proc_rank ( info->my_coords[0],
                              info->my_coords[1] + 1, info );
  xidx[iproc++] = xchg_count;

  for ( i = 1; i <= m; i++ )
  {
    recv_index[xchg_count]   = (m + 2) * (n + 1) + i;
    send_index[xchg_count++] = (m + 2) *  n      + i;
  }

  /* Upper-left cell. */

  rank[iproc]   = proc_rank ( info->my_coords[0] - 1,
                              info->my_coords[1] + 1, info );
  xidx[iproc++] = xchg_count;

  recv_index[xchg_count]   = (m + 2) * (n + 1);
  send_index[xchg_count++] = (m + 2) *  n + 1;

  /* Left border. */

  rank[iproc]   = proc_rank ( info->my_coords[0] - 1,
                              info->my_coords[1],     info );
  xidx[iproc++] = xchg_count;

  for ( j = 1; j <= n; j++ )
  {
    recv_index[xchg_count]   =     j * (m + 2);
    send_index[xchg_count++] = 1 + j * (m + 2);
  }

  xidx[iproc] = xchg_count;

  assert ( iproc      == 8 );
  assert ( xchg_count == (2 * m + 2 * n + 4) );

  xchg_count = 0;
  req_count  = 0;

  for ( iproc = 0; iproc < 8; iproc++ )
  {
    if ( rank[iproc] >= 0 )
    {
      i = xidx[iproc];
      j = xidx[iproc + 1];

      MPI_Send_init ( &send_buffer[xchg_count], (j - i), MPI_BYTE,
                      rank[iproc], EXCHANGE_TAG, MPI_COMM_WORLD,
                      &send_reqs[req_count] );

      MPI_Recv_init ( &recv_buffer[xchg_count], (j - i), MPI_BYTE,
                      rank[iproc], EXCHANGE_TAG, MPI_COMM_WORLD,
                      &recv_reqs[req_count] );

      req_count++;

      for ( ; i < j; i++, xchg_count++ )
      {
        recv_index[xchg_count] = recv_index[i];
        send_index[xchg_count] = send_index[i];
      }
    }
  }

  xdata->req_count   = req_count;
  xdata->xchg_count  = xchg_count;
  xdata->recv_buffer = recv_buffer;
  xdata->send_buffer = send_buffer;
  xdata->recv_index  = recv_index;
  xdata->send_index  = send_index;
  xdata->recv_reqs   = recv_reqs;
  xdata->send_reqs   = send_reqs;
}

/* ---------------------------------------------------------------------
 * exchange()
 * ------------------------------------------------------------------ */

void                      exchange

  ( grid_t*                 grid,
    xdata_t*                xdata )

{
  const int      xchg_count     = xdata->xchg_count;

  const int*     index          = NULL;
  const byte_t*  rhs            = NULL;
  byte_t*        lhs            = NULL;

  int            i;

  MPI_Startall ( xdata->req_count, xdata->recv_reqs );

  /* Load the send buffer. */

  index = xdata->send_index;
  lhs   = xdata->send_buffer;
  rhs   = grid->cells;

  for ( i = 0; i < xchg_count; i++ )
  {
    lhs[i] = rhs[index[i]];
  }

  MPI_Startall ( xdata->req_count, xdata->send_reqs );

  /* Wait for the send requests to complete and copy the receive
   * buffer to the grid. */

  index = xdata->recv_index;
  lhs   = grid->cells;
  rhs   = xdata->recv_buffer;

  MPI_Waitall ( xdata->req_count, xdata->recv_reqs,
                MPI_STATUSES_IGNORE );

  for ( i = 0; i < xchg_count; i++ )
  {
    lhs[index[i]] = rhs[i];
  }

  /* Wait for the send requests to complete. */

  MPI_Waitall ( xdata->req_count, xdata->send_reqs,
                MPI_STATUSES_IGNORE );
}

/* ---------------------------------------------------------------------
 * part_grid()
 * ------------------------------------------------------------------ */

void                      part_grid

  ( proc_info_t*            info,
    grid_t*                 grid,
    grid_t*                 grid0 )

{
  const int  m0    = grid0->global_dims[0];
  const int  n0    = grid0->global_dims[1];

  byte_t*    cells = NULL;
  byte_t*    row   = NULL;

  int        ibuf[10];
  int        count;
  int        rank;
  int        diff;
  int        mproc, nproc;
  int        iproc, jproc;
  int        i0,    i1;
  int        j0,    j1;
  int        m,     n;
  int        i,     j;
  int        k;

  MPI_Comm_size ( MPI_COMM_WORLD, &count );

  info->my_rank      = 0;
  info->my_coords[0] = 0;
  info->my_coords[1] = 0;

  /* Determine the optimal processor grid dimensions. */

  mproc = -1;
  nproc = -1;
  diff  = -1;

  for ( j = 1; j <= n0; j++ )
  {
    for ( i = 1; i <= m0; i++ )
    {
      if ( (i * j) == count )
      {
        if ( (diff < 0) || (abs( i - j ) < diff) )
        {
          mproc = i;
          nproc = j;
          diff  = abs ( i - j );
        }
      }
    }
  }

  if ( diff < 0 )
  {
    error ( "more processors than cells specified" );
  }

  info->proc_dims[0] = mproc;
  info->proc_dims[1] = nproc;

  if ( debug_level )
  {
    printf ( "Processor grid : %d x %d\n\n", mproc, nproc );
    printf ( "Distributing partitions ...\n" );
  }

  /* Partition the grid and distribute it to the other processes. */

  j0 = 0;

  for ( jproc = 0; jproc < nproc; jproc++ )
  {
    n  = (n0 - j0) / (nproc - jproc);
    j1 = j0 + n;
    i0 = 0;

    assert ( (j1 > j0) && (j1 <= n0) );

    for ( iproc = 0; iproc < mproc; iproc++ )
    {
      rank  = proc_rank ( iproc, jproc, info );
      m     = (m0 - i0) / (mproc - iproc);
      i1    = i0 + m;

      assert ( (i1 > i0) && (i1 <= m0) );

      cells = (byte_t*) allocate ( sizeof(*cells), (m + 2) * (n + 2) );
      k     = 0;

      for ( j = j0 - 1; j <= j1; j++ )
      {
        row = grid0->cells + ((j + 1) * (m0 + 2) + 1);

        for ( i = i0 - 1; i <= i1; i++ )
        {
          cells[k++] = row[i];
        }
      }

      if ( rank == 0 )
      {
        grid->origin     [0] = i0;
        grid->origin     [1] = j0;
        grid->local_dims [0] = m;
        grid->local_dims [1] = n;
        grid->global_dims[0] = m0;
        grid->global_dims[1] = n0;
        grid->cells          = cells;
      }
      else
      {
        ibuf[0] = iproc;
        ibuf[1] = jproc;
        ibuf[2] = mproc;
        ibuf[3] = nproc;
        ibuf[4] = i0;
        ibuf[5] = j0;
        ibuf[6] = m;
        ibuf[7] = n;
        ibuf[8] = m0;
        ibuf[9] = n0;

        MPI_Send ( ibuf, 10, MPI_INT, rank,
                   INIT_TAG, MPI_COMM_WORLD );

        MPI_Send ( cells, (m + 2) * (n + 2), MPI_BYTE, rank,
                   INIT_TAG, MPI_COMM_WORLD );

        free ( cells );
      }

      i0 = i1;
    }

    j0 = j1;
  }
}

/* ---------------------------------------------------------------------
 * recv_grid()
 * ------------------------------------------------------------------ */

void                      recv_grid

  ( proc_info_t*            info,
    grid_t*                 grid )

{
  const int   rank = 0;

  MPI_Status  status;
  int         ibuf[10];
  int         m, n;

  MPI_Recv ( ibuf, 10, MPI_INT, rank,
             INIT_TAG, MPI_COMM_WORLD, &status );

  info->my_coords  [0] = ibuf[0];
  info->my_coords  [1] = ibuf[1];
  info->proc_dims  [0] = ibuf[2];
  info->proc_dims  [1] = ibuf[3];
  grid->origin     [0] = ibuf[4];
  grid->origin     [1] = ibuf[5];
  grid->local_dims [0] = ibuf[6];
  grid->local_dims [1] = ibuf[7];
  grid->global_dims[0] = ibuf[8];
  grid->global_dims[1] = ibuf[9];

  m = grid->local_dims[0];
  n = grid->local_dims[1];

  grid->cells = (byte_t*) allocate ( sizeof(*grid->cells), (m + 2) * (n + 2) );

  MPI_Recv ( grid->cells, (m + 2) * (n + 2), MPI_BYTE, rank,
             INIT_TAG, MPI_COMM_WORLD, &status );

  if ( debug_level >= DEBUG_ALL )
  {
    printf     ( "\nGrid at process %d:\n\n", info->my_rank );
    print_grid ( grid );
  }
}

/* ---------------------------------------------------------------------
 * run_step()
 * ------------------------------------------------------------------ */

void                      run_step

  ( grid_t*                 grid,
    byte_t**                work )

{
  static const byte_t  TRANS_TABLE[9][2] =
  {
    { 0, 0 },  /* Cells with less than 2 live neighbors die. */
    { 0, 0 },  /* Cells with less than 2 live neighbors die. */
    { 0, 1 },  /* A live cell with 2 live neighbors lives. */
    { 1, 1 },  /* Cells with 3 live neighbors live. */
    { 0, 0 },  /* Cells with more than 3 live neighbors die. */
    { 0, 0 },  /* Cells with more than 3 live neighbors die. */
    { 0, 0 },  /* Cells with more than 3 live neighbors die. */
    { 0, 0 },  /* Cells with more than 3 live neighbors die. */
    { 0, 0 },  /* Cells with more than 3 live neighbors die. */
  };

  const int      m = grid->local_dims[0];
  const int      n = grid->local_dims[1];

  const byte_t*  rhs = NULL;
  byte_t*        lhs = NULL;
  byte_t*        tmp = NULL;

  int            i, j;
  byte_t         k;

  for ( j = 1; j <= n; j++ )
  {
    lhs = (*work)     + (j * (m + 2));
    rhs = grid->cells + (j * (m + 2));

    for ( i = 1; i <= m; i++ )
    {
      k      = (rhs[i - m - 3] +
                rhs[i - m - 2] +
                rhs[i - m - 1] +
                rhs[i - 1]     +
                rhs[i + 1]     +
                rhs[i + m + 1] +
                rhs[i + m + 2] +
                rhs[i + m + 3]);

      lhs[i] = TRANS_TABLE[k][rhs[i]];
    }
  }

  tmp         = grid->cells;
  grid->cells = *work;
  *work       = tmp;
}

/* ---------------------------------------------------------------------
 * run_game()
 * ------------------------------------------------------------------ */

void                      run_game

  ( xdata_t*                xdata,
    grid_t*                 grid,
    const conf_t*           conf )

{
  const int  m = grid->local_dims[0];
  const int  n = grid->local_dims[1];

  byte_t*    work;
  int        istep;

  if ( debug_level )
  {
    printf ( "Starting simulation ...\n" );

    if ( debug_level >= DEBUG_ALL )
    {
      printf ( "\n" );
    }
  }

  work = (byte_t*) allocate ( sizeof(*work), (m + 2) * (n + 2) );

  for ( istep = 0; istep < conf->steps; istep++ )
  {
    exchange ( grid, xdata );
    run_step ( grid, &work );

    if ( debug_level >= DEBUG_ALL )
    {
      printf     ( "\nStep %d\n\n", istep );
      print_grid ( grid );
      usleep     ( 400 * 1000 );
    }
  }

  free ( work );
}

/* ---------------------------------------------------------------------
 * run_worker()
 * ------------------------------------------------------------------ */

void                      run_worker

  ( int                     my_rank )

{
  proc_info_t  info;
  conf_t       conf;
  grid_t       grid;
  xdata_t      xdata;

  memset ( &info,  0x0, sizeof(info) );
  memset ( &conf,  0x0, sizeof(conf) );
  memset ( &grid,  0x0, sizeof(grid) );
  memset ( &xdata, 0x0, sizeof(xdata) );

  info.my_rank = my_rank;

  recv_conf  ( &conf );
  recv_grid  ( &info,  &grid );
  init_xdata ( &xdata, &grid, &info );
  run_game   ( &xdata, &grid, &conf );
  kill_grid  ( &grid );
  kill_xdata ( &xdata );
}

/* ---------------------------------------------------------------------
 * run_master()
 * ------------------------------------------------------------------ */

void                      run_master

  ( int                     argc,
    char**                  argv )

{
  proc_info_t  info;
  conf_t       conf;
  grid_t       grid0;
  grid_t       grid;
  xdata_t      xdata;
  double       time0;

  memset ( &info,  0x0, sizeof(info) );
  memset ( &conf,  0x0, sizeof(conf) );
  memset ( &grid0, 0x0, sizeof(grid0) );
  memset ( &grid,  0x0, sizeof(grid) );
  memset ( &xdata, 0x0, sizeof(xdata) );

  time0 = MPI_Wtime ();

  parse_args ( &conf,  argc,  argv );
  init_grid  ( &grid0, &conf );
  part_grid  ( &info,  &grid, &grid0 );
  kill_grid  ( &grid0 );
  init_xdata ( &xdata, &grid, &info );
  run_game   ( &xdata, &grid, &conf );
  kill_grid  ( &grid );
  kill_xdata ( &xdata );

  printf ( "Finished in %g seconds.\n\n", (MPI_Wtime() - time0) );
}

/* ---------------------------------------------------------------------
 * main()
 * ------------------------------------------------------------------ */

int                       main

  ( int                     argc,
    char**                  argv )

{
  int  my_rank;

  MPI_Init      ( &argc, &argv );
  MPI_Comm_rank ( MPI_COMM_WORLD, &my_rank );

  srand ( my_rank );

  if ( my_rank == 0 )
  {
    run_master ( argc, argv );
  }
  else
  {
    run_worker ( my_rank );
  }

  MPI_Finalize ();

  return 0;
}
Câu hỏi 2: Number of MPI processes

Specify the number of MPI processes (1-64) that should be used for running your program (mpirun -np <nproc>)
Câu hỏi 3: Command Line Argument

Specify the command line arguments that should be passed to your program when it is run. Arguments must be given in quotes and separated by commas. Leave this filed empty if you do not want to specify command line arguments.
Example: "1","int","arg=2" is interpreted as three arguments 1, int, and arg=2.
Tác giả	Erik-Jan Lingen, Matthias Möller
Hạn chót	Không có hạn chót
Giới hạn nộp bài	Không có giới hạn
Thông tin

Đăng nhập
