Example #1
0
int main(int argc, char **argv) {
  // OP initialisation
  op_init(argc, argv, 2);

  // MPI for user I/O
  int my_rank;
  int comm_size;
  MPI_Comm_rank(MPI_COMM_WORLD, &my_rank);
  MPI_Comm_size(MPI_COMM_WORLD, &comm_size);

  // timer
  double cpu_t1, cpu_t2, wall_t1, wall_t2;

  int *pp;
  float *A, *r, *u, *du;

  int nnode, nedge;

  /**------------------------BEGIN I/O and PARTITIONING ---------------------**/

  int g_nnode, g_nedge, g_n, g_e;

  g_nnode = (NN - 1) * (NN - 1);
  g_nedge = (NN - 1) * (NN - 1) + 4 * (NN - 1) * (NN - 2);

  int *g_pp = 0;
  float *g_A = 0, *g_r = 0, *g_u = 0, *g_du = 0;

  op_printf("Global number of nodes, edges = %d, %d\n", g_nnode, g_nedge);

  if (my_rank == MPI_ROOT) {
    g_pp = (int *)malloc(sizeof(int) * 2 * g_nedge);

    g_A = (float *)malloc(sizeof(float) * g_nedge);
    g_r = (float *)malloc(sizeof(float) * g_nnode);
    g_u = (float *)malloc(sizeof(float) * g_nnode);
    g_du = (float *)malloc(sizeof(float) * g_nnode);

    // create matrix and r.h.s., and set coordinates needed for renumbering /
    // partitioning

    g_e = 0;

    for (int i = 1; i < NN; i++) {
      for (int j = 1; j < NN; j++) {
        g_n = i - 1 + (j - 1) * (NN - 1);
        g_r[g_n] = 0.0f;
        g_u[g_n] = 0.0f;
        g_du[g_n] = 0.0f;

        g_pp[2 * g_e] = g_n;
        g_pp[2 * g_e + 1] = g_n;
        g_A[g_e] = -1.0f;
        g_e++;

        for (int pass = 0; pass < 4; pass++) {
          int i2 = i;
          int j2 = j;
          if (pass == 0)
            i2 += -1;
          if (pass == 1)
            i2 += 1;
          if (pass == 2)
            j2 += -1;
          if (pass == 3)
            j2 += 1;

          if ((i2 == 0) || (i2 == NN) || (j2 == 0) || (j2 == NN)) {
            g_r[g_n] += 0.25f;
          } else {
            g_pp[2 * g_e] = g_n;
            g_pp[2 * g_e + 1] = i2 - 1 + (j2 - 1) * (NN - 1);
            g_A[g_e] = 0.25f;
            g_e++;
          }
        }
      }
    }
  }

  /* Compute local sizes */
  nnode = compute_local_size(g_nnode, comm_size, my_rank);
  nedge = compute_local_size(g_nedge, comm_size, my_rank);
  op_printf("Number of nodes, edges on process %d = %d, %d\n", my_rank, nnode,
            nedge);

  /*Allocate memory to hold local sets, mapping tables and data*/
  pp = (int *)malloc(2 * sizeof(int) * nedge);

  A = (float *)malloc(nedge * sizeof(float));
  r = (float *)malloc(nnode * sizeof(float));
  u = (float *)malloc(nnode * sizeof(float));
  du = (float *)malloc(nnode * sizeof(float));

  /* scatter sets, mappings and data on sets*/
  scatter_int_array(g_pp, pp, comm_size, g_nedge, nedge, 2);
  scatter_float_array(g_A, A, comm_size, g_nedge, nedge, 1);
  scatter_float_array(g_r, r, comm_size, g_nnode, nnode, 1);
  scatter_float_array(g_u, u, comm_size, g_nnode, nnode, 1);
  scatter_float_array(g_du, du, comm_size, g_nnode, nnode, 1);

  /*Freeing memory allocated to gloabal arrays on rank 0
    after scattering to all processes*/
  if (my_rank == MPI_ROOT) {
    free(g_pp);
    free(g_A);
    free(g_r);
    free(g_u);
    free(g_du);
  }

  /**------------------------END I/O and PARTITIONING ---------------------**/

  // declare sets, pointers, and datasets

  op_set nodes = op_decl_set(nnode, "nodes");
  op_set edges = op_decl_set(nedge, "edges");

  op_map ppedge = op_decl_map(edges, nodes, 2, pp, "ppedge");

  op_dat p_A = op_decl_dat(edges, 1, "float", A, "p_A");
  op_dat p_r = op_decl_dat(nodes, 1, "float", r, "p_r");
  op_dat p_u = op_decl_dat(nodes, 1, "float", u, "p_u");
  op_dat p_du = op_decl_dat(nodes, 1, "float", du, "p_du");

  alpha = 1.0f;
  op_decl_const(1, "float", &alpha);

  op_diagnostic_output();

  // trigger partitioning and halo creation routines
  op_partition("PTSCOTCH", "KWAY", NULL, NULL, NULL);

  // initialise timers for total execution wall time
  op_timers(&cpu_t1, &wall_t1);

  // main iteration loop

  float u_sum, u_max, beta = 1.0f;

  for (int iter = 0; iter < NITER; iter++) {
    op_par_loop(res, "res", edges,
                op_arg_dat(p_A, -1, OP_ID, 1, "float", OP_READ),
                op_arg_dat(p_u, 1, ppedge, 1, "float", OP_READ),
                op_arg_dat(p_du, 0, ppedge, 1, "float", OP_INC),
                op_arg_gbl(&beta, 1, "float", OP_READ));

    u_sum = 0.0f;
    u_max = 0.0f;
    op_par_loop(update, "update", nodes,
                op_arg_dat(p_r, -1, OP_ID, 1, "float", OP_READ),
                op_arg_dat(p_du, -1, OP_ID, 1, "float", OP_RW),
                op_arg_dat(p_u, -1, OP_ID, 1, "float", OP_INC),
                op_arg_gbl(&u_sum, 1, "float", OP_INC),
                op_arg_gbl(&u_max, 1, "float", OP_MAX));

    op_printf("\n u max/rms = %f %f \n\n", u_max, sqrt(u_sum / g_nnode));
  }

  op_timers(&cpu_t2, &wall_t2);

  // get results data array
  op_fetch_data(p_u, u);

  // output the result dat array to files
  op_print_dat_to_txtfile(p_u, "out_grid_mpi.dat"); // ASCI
  op_print_dat_to_binfile(p_u, "out_grid_mpi.bin"); // Binary

  printf("solution on rank %d\n", my_rank);
  for (int i = 0; i < nnode; i++) {
    printf(" %7.4f", u[i]);
    fflush(stdout);
  }
  printf("\n");

  // print each mpi process's timing info for each kernel
  op_timing_output();

  // print total time for niter interations
  op_printf("Max total runtime = %f\n", wall_t2 - wall_t1);

  // gather results from all ranks and check
  float *ug = (float *)malloc(sizeof(float) * op_get_size(nodes));
  op_fetch_data_idx(p_u, ug, 0, op_get_size(nodes) - 1);
  int result = check_result<float>(ug, NN, TOLERANCE);
  free(ug);

  op_exit();

  free(u);
  free(pp);
  free(A);
  free(r);
  free(du);

  return result;
}
Example #2
0
int main(int argc, char **argv)
{
  // OP initialisation
  op_init(argc,argv,2);

  //MPI for user I/O
  int my_rank;
  int comm_size;
  MPI_Comm_rank(MPI_COMM_WORLD, &my_rank);
  MPI_Comm_size(MPI_COMM_WORLD, &comm_size);

  //timer
  double cpu_t1, cpu_t2, wall_t1, wall_t2;

  int    *becell, *ecell,  *bound, *bedge, *edge, *cell;
  double  *x, *q, *qold, *adt, *res;

  int    nnode,ncell,nedge,nbedge,niter;
  double  rms;

  /**------------------------BEGIN I/O and PARTITIONING -------------------**/

  op_timers(&cpu_t1, &wall_t1);

  /* read in grid from disk on root processor */
  FILE *fp;

  if ( (fp = fopen("new_grid.dat","r")) == NULL) {
    op_printf("can't open file new_grid.dat\n"); exit(-1);
  }

  int   g_nnode,g_ncell,g_nedge,g_nbedge;

  check_scan(fscanf(fp,"%d %d %d %d \n",&g_nnode, &g_ncell, &g_nedge, &g_nbedge), 4);

  int *g_becell = 0, *g_ecell = 0, *g_bound = 0, *g_bedge = 0, *g_edge = 0, *g_cell = 0;
  double *g_x = 0,*g_q = 0, *g_qold = 0, *g_adt = 0, *g_res = 0;

  // set constants

  op_printf("initialising flow field\n");
  gam = 1.4f;
  gm1 = gam - 1.0f;
  cfl = 0.9f;
  eps = 0.05f;

  double mach  = 0.4f;
  double alpha = 3.0f*atan(1.0f)/45.0f;
  double p     = 1.0f;
  double r     = 1.0f;
  double u     = sqrt(gam*p/r)*mach;
  double e     = p/(r*gm1) + 0.5f*u*u;

  qinf[0] = r;
  qinf[1] = r*u;
  qinf[2] = 0.0f;
  qinf[3] = r*e;

  op_printf("reading in grid \n");
  op_printf("Global number of nodes, cells, edges, bedges = %d, %d, %d, %d\n"
      ,g_nnode,g_ncell,g_nedge,g_nbedge);

  if(my_rank == MPI_ROOT) {
    g_cell   = (int *) malloc(4*g_ncell*sizeof(int));
    g_edge   = (int *) malloc(2*g_nedge*sizeof(int));
    g_ecell  = (int *) malloc(2*g_nedge*sizeof(int));
    g_bedge  = (int *) malloc(2*g_nbedge*sizeof(int));
    g_becell = (int *) malloc(  g_nbedge*sizeof(int));
    g_bound  = (int *) malloc(  g_nbedge*sizeof(int));

    g_x      = (double *) malloc(2*g_nnode*sizeof(double));
    g_q      = (double *) malloc(4*g_ncell*sizeof(double));
    g_qold   = (double *) malloc(4*g_ncell*sizeof(double));
    g_res    = (double *) malloc(4*g_ncell*sizeof(double));
    g_adt    = (double *) malloc(  g_ncell*sizeof(double));

    for (int n=0; n<g_nnode; n++){
      check_scan(fscanf(fp,"%lf %lf \n",&g_x[2*n], &g_x[2*n+1]), 2);
    }

    for (int n=0; n<g_ncell; n++) {
      check_scan(fscanf(fp,"%d %d %d %d \n",&g_cell[4*n  ], &g_cell[4*n+1],
            &g_cell[4*n+2], &g_cell[4*n+3]), 4);
    }

    for (int n=0; n<g_nedge; n++) {
      check_scan(fscanf(fp,"%d %d %d %d \n",&g_edge[2*n],&g_edge[2*n+1],
            &g_ecell[2*n],&g_ecell[2*n+1]), 4);
    }

    for (int n=0; n<g_nbedge; n++) {
      check_scan(fscanf(fp,"%d %d %d %d \n",&g_bedge[2*n],&g_bedge[2*n+1],
            &g_becell[n],&g_bound[n]), 4);
    }

    //initialise flow field and residual

    for (int n=0; n<g_ncell; n++) {
      for (int m=0; m<4; m++) {
        g_q[4*n+m] = qinf[m];
        g_res[4*n+m] = 0.0f;
      }
    }
  }

  fclose(fp);

  nnode = compute_local_size (g_nnode, comm_size, my_rank);
  ncell = compute_local_size (g_ncell, comm_size, my_rank);
  nedge = compute_local_size (g_nedge, comm_size, my_rank);
  nbedge = compute_local_size (g_nbedge, comm_size, my_rank);

  op_printf("Number of nodes, cells, edges, bedges on process %d = %d, %d, %d, %d\n"
      ,my_rank,nnode,ncell,nedge,nbedge);

  /*Allocate memory to hold local sets, mapping tables and data*/
  cell   = (int *) malloc(4*ncell*sizeof(int));
  edge   = (int *) malloc(2*nedge*sizeof(int));
  ecell  = (int *) malloc(2*nedge*sizeof(int));
  bedge  = (int *) malloc(2*nbedge*sizeof(int));
  becell = (int *) malloc(  nbedge*sizeof(int));
  bound  = (int *) malloc(  nbedge*sizeof(int));

  x      = (double *) malloc(2*nnode*sizeof(double));
  q      = (double *) malloc(4*ncell*sizeof(double));
  qold   = (double *) malloc(4*ncell*sizeof(double));
  res    = (double *) malloc(4*ncell*sizeof(double));
  adt    = (double *) malloc(  ncell*sizeof(double));

  /* scatter sets, mappings and data on sets*/
  scatter_int_array(g_cell, cell, comm_size, g_ncell,ncell, 4);
  scatter_int_array(g_edge, edge, comm_size, g_nedge,nedge, 2);
  scatter_int_array(g_ecell, ecell, comm_size, g_nedge,nedge, 2);
  scatter_int_array(g_bedge, bedge, comm_size, g_nbedge,nbedge, 2);
  scatter_int_array(g_becell, becell, comm_size, g_nbedge,nbedge, 1);
  scatter_int_array(g_bound, bound, comm_size, g_nbedge,nbedge, 1);

  scatter_double_array(g_x, x, comm_size, g_nnode,nnode, 2);
  scatter_double_array(g_q, q, comm_size, g_ncell,ncell, 4);
  scatter_double_array(g_qold, qold, comm_size, g_ncell,ncell, 4);
  scatter_double_array(g_res, res, comm_size, g_ncell,ncell, 4);
  scatter_double_array(g_adt, adt, comm_size, g_ncell,ncell, 1);

  /*Freeing memory allocated to gloabal arrays on rank 0
    after scattering to all processes*/
  if(my_rank == MPI_ROOT) {
    free(g_cell);
    free(g_edge);
    free(g_ecell);
    free(g_bedge);
    free(g_becell);
    free(g_bound);
    free(g_x );
    free(g_q);
    free(g_qold);
    free(g_adt);
    free(g_res);
  }

  op_timers(&cpu_t2, &wall_t2);
  op_printf("Max total file read time = %f\n", wall_t2-wall_t1);

  /**------------------------END I/O and PARTITIONING -----------------------**/

  // declare sets, pointers, datasets and global constants

  op_set nodes  = op_decl_set(nnode,  "nodes");
  op_set edges  = op_decl_set(nedge,  "edges");
  op_set bedges = op_decl_set(nbedge, "bedges");
  op_set cells  = op_decl_set(ncell,  "cells");

  op_map pedge   = op_decl_map(edges, nodes,2,edge,  "pedge");
  op_map pecell  = op_decl_map(edges, cells,2,ecell, "pecell");
  op_map pbedge  = op_decl_map(bedges,nodes,2,bedge, "pbedge");
  op_map pbecell = op_decl_map(bedges,cells,1,becell,"pbecell");
  op_map pcell   = op_decl_map(cells, nodes,4,cell,  "pcell");

  op_dat p_bound = op_decl_dat(bedges,1,"int"  ,bound,"p_bound");
  op_dat p_x     = op_decl_dat(nodes ,2,"double",x    ,"p_x");
  op_dat p_q     = op_decl_dat(cells ,4,"double",q    ,"p_q");
  op_dat p_qold  = op_decl_dat(cells ,4,"double",qold ,"p_qold");
  op_dat p_adt   = op_decl_dat(cells ,1,"double",adt  ,"p_adt");
  op_dat p_res   = op_decl_dat(cells ,4,"double",res  ,"p_res");

  op_decl_const2("gam",1,"double",&gam);
  op_decl_const2("gm1",1,"double",&gm1);
  op_decl_const2("cfl",1,"double",&cfl);
  op_decl_const2("eps",1,"double",&eps);
  op_decl_const2("mach",1,"double",&mach);
  op_decl_const2("alpha",1,"double",&alpha);
  op_decl_const2("qinf",4,"double",qinf);

  op_diagnostic_output();

  //trigger partitioning and halo creation routines
  op_partition("PTSCOTCH", "KWAY", cells, pecell, p_x);
  //op_partition("PARMETIS", "KWAY", cells, pecell, p_x);

  //initialise timers for total execution wall time
  op_timers(&cpu_t1, &wall_t1);

  niter = 1000;
  for(int iter=1; iter<=niter; iter++) {

    //save old flow solution
    op_par_loop_save_soln("save_soln",cells,
                op_arg_dat(p_q,-1,OP_ID,4,"double",OP_READ),
                op_arg_dat(p_qold,-1,OP_ID,4,"double",OP_WRITE));

    //  predictor/corrector update loop

    for(int k=0; k<2; k++) {

      //    calculate area/timstep
      op_par_loop_adt_calc("adt_calc",cells,
                  op_arg_dat(p_x,0,pcell,2,"double",OP_READ),
                  op_arg_dat(p_x,1,pcell,2,"double",OP_READ),
                  op_arg_dat(p_x,2,pcell,2,"double",OP_READ),
                  op_arg_dat(p_x,3,pcell,2,"double",OP_READ),
                  op_arg_dat(p_q,-1,OP_ID,4,"double",OP_READ),
                  op_arg_dat(p_adt,-1,OP_ID,1,"double",OP_WRITE));

      //    calculate flux residual
      op_par_loop_res_calc("res_calc",edges,
                  op_arg_dat(p_x,0,pedge,2,"double",OP_READ),
                  op_arg_dat(p_x,1,pedge,2,"double",OP_READ),
                  op_arg_dat(p_q,0,pecell,4,"double",OP_READ),
                  op_arg_dat(p_q,1,pecell,4,"double",OP_READ),
                  op_arg_dat(p_adt,0,pecell,1,"double",OP_READ),
                  op_arg_dat(p_adt,1,pecell,1,"double",OP_READ),
                  op_arg_dat(p_res,0,pecell,4,"double",OP_INC),
                  op_arg_dat(p_res,1,pecell,4,"double",OP_INC));

      op_par_loop_bres_calc("bres_calc",bedges,
                  op_arg_dat(p_x,0,pbedge,2,"double",OP_READ),
                  op_arg_dat(p_x,1,pbedge,2,"double",OP_READ),
                  op_arg_dat(p_q,0,pbecell,4,"double",OP_READ),
                  op_arg_dat(p_adt,0,pbecell,1,"double",OP_READ),
                  op_arg_dat(p_res,0,pbecell,4,"double",OP_INC),
                  op_arg_dat(p_bound,-1,OP_ID,1,"int",OP_READ));

      //    update flow field

      rms = 0.0;

      op_par_loop_update("update",cells,
                  op_arg_dat(p_qold,-1,OP_ID,4,"double",OP_READ),
                  op_arg_dat(p_q,-1,OP_ID,4,"double",OP_WRITE),
                  op_arg_dat(p_res,-1,OP_ID,4,"double",OP_RW),
                  op_arg_dat(p_adt,-1,OP_ID,1,"double",OP_READ),
                  op_arg_gbl(&rms,1,"double",OP_INC));
    }

    //print iteration history
    rms = sqrt(rms/(double) g_ncell);
    if (iter%100 == 0)
      op_printf("%d  %10.5e \n",iter,rms);
  }

  op_timers(&cpu_t2, &wall_t2);

  //output the result dat array to files
  op_print_dat_to_txtfile(p_q, "out_grid_mpi.dat"); //ASCI
  op_print_dat_to_binfile(p_q, "out_grid_mpi.bin"); //Binary

  //write given op_dat's indicated segment of data to a memory block in the order it was originally
  //arranged (i.e. before partitioning and reordering)
  double* q_part = (double *)op_malloc(sizeof(double)*op_get_size(cells)*4);
  op_fetch_data_idx(p_q, q_part, 0, op_get_size(cells)-1);
  free(q_part);

  op_timing_output();
  op_printf("Max total runtime = %f\n",wall_t2-wall_t1);

  op_exit();

  free(cell);
  free(edge);
  free(ecell);
  free(bedge);
  free(becell);
  free(bound);
  free(x);
  free(q);
  free(qold);
  free(res);
  free(adt);
}
Example #3
0
int main(int argc, char **argv)
{
  // OP initialisation
  op_init(argc,argv,2);

  int    *becell, *ecell,  *bound, *bedge, *edge, *cell;
  double  *x, *q, *qold, *adt, *res;

  int    nnode,ncell,nedge,nbedge,niter;
  double  rms;

  //timer
  double cpu_t1, cpu_t2, wall_t1, wall_t2;

  // read in grid

  op_printf("reading in grid \n");

  FILE *fp;
  if ( (fp = fopen("./new_grid.dat","r")) == NULL) {
    op_printf("can't open file new_grid.dat\n"); exit(-1);
  }

  if (fscanf(fp,"%d %d %d %d \n",&nnode, &ncell, &nedge, &nbedge) != 4) {
    op_printf("error reading from new_grid.dat\n"); exit(-1);
  }

  cell   = (int *) malloc(4*ncell*sizeof(int));
  edge   = (int *) malloc(2*nedge*sizeof(int));
  ecell  = (int *) malloc(2*nedge*sizeof(int));
  bedge  = (int *) malloc(2*nbedge*sizeof(int));
  becell = (int *) malloc(  nbedge*sizeof(int));
  bound  = (int *) malloc(  nbedge*sizeof(int));

  x      = (double *) malloc(2*nnode*sizeof(double));
  q      = (double *) malloc(4*ncell*sizeof(double));
  qold   = (double *) malloc(4*ncell*sizeof(double));
  res    = (double *) malloc(4*ncell*sizeof(double));
  adt    = (double *) malloc(  ncell*sizeof(double));

  for (int n=0; n<nnode; n++) {
    if (fscanf(fp,"%lf %lf \n",&x[2*n], &x[2*n+1]) != 2) {
      op_printf("error reading from new_grid.dat\n"); exit(-1);
    }
  }

  for (int n=0; n<ncell; n++) {
    if (fscanf(fp,"%d %d %d %d \n",&cell[4*n  ], &cell[4*n+1],
                                   &cell[4*n+2], &cell[4*n+3]) != 4) {
      op_printf("error reading from new_grid.dat\n"); exit(-1);
    }
  }

  for (int n=0; n<nedge; n++) {
    if (fscanf(fp,"%d %d %d %d \n",&edge[2*n], &edge[2*n+1],
                                   &ecell[2*n],&ecell[2*n+1]) != 4) {
      op_printf("error reading from new_grid.dat\n"); exit(-1);
    }
  }

  for (int n=0; n<nbedge; n++) {
    if (fscanf(fp,"%d %d %d %d \n",&bedge[2*n],&bedge[2*n+1],
                                   &becell[n], &bound[n]) != 4) {
      op_printf("error reading from new_grid.dat\n"); exit(-1);
    }
  }

  fclose(fp);

  // set constants and initialise flow field and residual

  op_printf("initialising flow field \n");

  gam = 1.4f;
  gm1 = gam - 1.0f;
  cfl = 0.9f;
  eps = 0.05f;

  double mach  = 0.4f;
  double alpha = 3.0f*atan(1.0f)/45.0f;
  double p     = 1.0f;
  double r     = 1.0f;
  double u     = sqrt(gam*p/r)*mach;
  double e     = p/(r*gm1) + 0.5f*u*u;

  qinf[0] = r;
  qinf[1] = r*u;
  qinf[2] = 0.0f;
  qinf[3] = r*e;

  for (int n=0; n<ncell; n++) {
    for (int m=0; m<4; m++) {
        q[4*n+m] = qinf[m];
      res[4*n+m] = 0.0f;
    }
  }

  // declare sets, pointers, datasets and global constants

  op_set nodes  = op_decl_set(nnode,  "nodes");
  op_set edges  = op_decl_set(nedge,  "edges");
  op_set bedges = op_decl_set(nbedge, "bedges");
  op_set cells  = op_decl_set(ncell,  "cells");

  op_map pedge   = op_decl_map(edges, nodes,2,edge,  "pedge");
  op_map pecell  = op_decl_map(edges, cells,2,ecell, "pecell");
  op_map pbedge  = op_decl_map(bedges,nodes,2,bedge, "pbedge");
  op_map pbecell = op_decl_map(bedges,cells,1,becell,"pbecell");
  op_map pcell   = op_decl_map(cells, nodes,4,cell,  "pcell");

  op_dat p_bound = op_decl_dat(bedges,1,"int"  ,bound,"p_bound");
  op_dat p_x     = op_decl_dat(nodes ,2,"double",x    ,"p_x");
  op_dat p_q     = op_decl_dat(cells ,4,"double",q    ,"p_q");
  op_dat p_qold  = op_decl_dat(cells ,4,"double",qold ,"p_qold");
  op_dat p_adt   = op_decl_dat(cells ,1,"double",adt  ,"p_adt");
  op_dat p_res   = op_decl_dat(cells ,4,"double",res  ,"p_res");

  op_decl_const(1,"double",&gam  );
  op_decl_const(1,"double",&gm1  );
  op_decl_const(1,"double",&cfl  );
  op_decl_const(1,"double",&eps  );
  op_decl_const(1,"double",&mach );
  op_decl_const(1,"double",&alpha);
  op_decl_const(4,"double",qinf  );

  op_diagnostic_output();

  //initialise timers for total execution wall time
  op_timers(&cpu_t1, &wall_t1);

  // main time-marching loop

  niter = 1000;

  for(int iter=1; iter<=niter; iter++) {

    // save old flow solution

    op_par_loop(save_soln,"save_soln", cells,
      op_arg_dat(p_q,   -1,OP_ID, 4,"double",OP_READ ),
      op_arg_dat(p_qold,-1,OP_ID, 4,"double",OP_WRITE));

    // predictor/corrector update loop

    for(int k=0; k<2; k++) {

      // calculate area/timstep

      op_par_loop(adt_calc,"adt_calc",cells,
          op_arg_dat(p_x,   0,pcell, 2,"double",OP_READ ),
          op_arg_dat(p_x,   1,pcell, 2,"double",OP_READ ),
          op_arg_dat(p_x,   2,pcell, 2,"double",OP_READ ),
          op_arg_dat(p_x,   3,pcell, 2,"double",OP_READ ),
          op_arg_dat(p_q,  -1,OP_ID, 4,"double",OP_READ ),
          op_arg_dat(p_adt,-1,OP_ID, 1,"double",OP_WRITE));

      // calculate flux residual

      op_par_loop(res_calc,"res_calc",edges,
          op_arg_dat(p_x,    0,pedge, 2,"double",OP_READ),
          op_arg_dat(p_x,    1,pedge, 2,"double",OP_READ),
          op_arg_dat(p_q,    0,pecell,4,"double",OP_READ),
          op_arg_dat(p_q,    1,pecell,4,"double",OP_READ),
          op_arg_dat(p_adt,  0,pecell,1,"double",OP_READ),
          op_arg_dat(p_adt,  1,pecell,1,"double",OP_READ),
          op_arg_dat(p_res,  0,pecell,4,"double",OP_INC ),
          op_arg_dat(p_res,  1,pecell,4,"double",OP_INC ));

      op_par_loop(bres_calc,"bres_calc",bedges,
          op_arg_dat(p_x,     0,pbedge, 2,"double",OP_READ),
          op_arg_dat(p_x,     1,pbedge, 2,"double",OP_READ),
          op_arg_dat(p_q,     0,pbecell,4,"double",OP_READ),
          op_arg_dat(p_adt,   0,pbecell,1,"double",OP_READ),
          op_arg_dat(p_res,   0,pbecell,4,"double",OP_INC ),
          op_arg_dat(p_bound,-1,OP_ID  ,1,"int",  OP_READ));

      // update flow field

      rms = 0.0;

      op_par_loop(update,"update",cells,
          op_arg_dat(p_qold,-1,OP_ID, 4,"double",OP_READ ),
          op_arg_dat(p_q,   -1,OP_ID, 4,"double",OP_WRITE),
          op_arg_dat(p_res, -1,OP_ID, 4,"double",OP_RW   ),
          op_arg_dat(p_adt, -1,OP_ID, 1,"double",OP_READ ),
          op_arg_gbl(&rms,1,"double",OP_INC));
    }

    // print iteration history
    rms = sqrt(rms/(double) op_get_size(cells));
    if (iter%100 == 0)
      op_printf(" %d  %10.5e \n",iter,rms);
  }

  op_timers(&cpu_t2, &wall_t2);

  //output the result dat array to files
  op_print_dat_to_txtfile(p_q, "out_grid_seq.dat"); //ASCI
  op_print_dat_to_binfile(p_q, "out_grid_seq.bin"); //Binary

  op_timing_output();
  op_printf("Max total runtime = \n%f\n",wall_t2-wall_t1);

  op_exit();

  free(cell);
  free(edge);
  free(ecell);
  free(bedge);
  free(becell);
  free(bound);
  free(x);
  free(q);
  free(qold);
  free(res);
  free(adt);
}