void ops_dat_fetch_data(ops_dat dat, int part, char *data) {
ops_get_data(dat);
int lsize[OPS_MAX_DIM] = {1};
int ldisp[OPS_MAX_DIM] = {1};
ops_dat_get_extents(dat, part, ldisp, lsize);
lsize[0] *= dat->elem_size/dat->dim; //now in bytes
if (dat->block->dims>3) {ops_printf("Error, ops_dat_fetch_data not implemented for dims>3\n"); exit(-1);}
if (OPS_soa && dat->dim > 1) {ops_printf("Error, ops_dat_fetch_data not implemented for SoA\n"); exit(-1);}
for (int k = 0; k < lsize[2]; k++)
for (int j = 0; j < lsize[1]; j++)
memcpy(&data[k*lsize[0]*lsize[1]+j*lsize[0]],
&dat->data[((j-dat->d_m[1] + (k-dat->d_m[2])*dat->size[1])*dat->size[0] - dat->d_m[0])* dat->elem_size],
lsize[0]);
}
void tea_leaf_cheby_first_step(double *ch_alphas, double *ch_betas, int *fields,
double *error, double *theta, double cn, int max_cheby_iters, int *est_itc, double solve_time, double rx, double ry) {
double bb = 0;
// calculate 2 norm of u0
tea_leaf_calc_2norm(0, &bb);
// initialise 'p' array
tea_leaf_cheby_init(u,u0,vector_p,vector_r,vector_Mi,vector_w,vector_z,vector_Kx,vector_Ky,tri_cp,tri_bfp,rx,ry,*theta,tl_preconditioner_type);
// if (profiler_on) halo_time = timer()
update_halo(fields,1);
// if (profiler_on) solve_time = solve_time + (timer()-halo_time)
tea_leaf_cheby_iterate(u,u0,vector_p,vector_r,vector_Mi,vector_w,vector_z,vector_Kx,vector_Ky,tri_cp,tri_bfp,ch_alphas, ch_betas, rx,ry,1,tl_preconditioner_type);
tea_leaf_calc_2norm(1, error);
double it_alpha = eps/2.0*sqrt(bb/(*error));//eps*bb/(4.0*(*error));
double gamm = (sqrt(cn) - 1.0)/(sqrt(cn) + 1.0);
*est_itc = round(log(it_alpha)/(log(gamm)));
ops_fprintf(g_out," est itc\n%11d\n",*est_itc);
ops_printf(" est itc\n%11d\n",*est_itc);
}
int main(int argc, const char **argv) {
ops_init(argc, argv, 5);
ops_init_backend();
ops_printf("Hello world from OPS!\n\n");
ops_block block = ops_decl_block_hdf5(3, "grid0", "write_data.h5");
ops_dat single =
ops_decl_dat_hdf5(block, 1, "double", "single", "write_data.h5");
ops_dat multi =
ops_decl_dat_hdf5(block, 2, "double", "multi", "write_data.h5");
ops_dat integ = ops_decl_dat_hdf5(block, 1, "int", "integ", "write_data.h5");
ops_partition("empty_string_that_does_nothing_yet");
ops_diagnostic_output();
ops_fetch_block_hdf5_file(block, "read_data.h5");
ops_fetch_dat_hdf5_file(multi, "read_data.h5");
ops_fetch_dat_hdf5_file(single, "read_data.h5");
ops_fetch_dat_hdf5_file(integ, "read_data.h5");
int my_const;
ops_get_const_hdf5("my_const", 1, "int", (char *)&my_const, "write_data.h5");
printf("Read const: %d\n", my_const);
char buffer[50];
ops_get_const_hdf5("my_text", 11, "char", buffer, "write_data.h5");
printf("Read text: %s\n", buffer);
ops_write_const_hdf5("my_const", 1, "int", (char *)&my_const, "read_data.h5");
ops_write_const_hdf5("my_text", 11, "char", (char *)buffer, "read_data.h5");
ops_timing_output(stdout);
ops_printf("\nSucessful exit from OPS!\n");
ops_exit();
}
void timestep() {
double dtlp;
double kernel_time, c, t;
if(profiler_on) ops_timers_core(&c,&kernel_time);
//calc_dt(&dt)
dt = dtinit;
// if(profiler_on) profiler%timestep=profiler%timestep+(timer()-kernel_time)
ops_fprintf(g_out, " Step %8d time %.7lf timestep %-10.2E\n", step, currtime, dt);
ops_printf(" Step %8d time %.7lf timestep %-10.2E\n", step, currtime, dt);
}
void field_summary()
{
double qa_diff;
//initialize sizes using global values
int x_min = field.x_min;
int x_max = field.x_max;
int y_min = field.y_min;
int y_max = field.y_max;
int rangexy_inner[] = {x_min,x_max,y_min,y_max}; // inner range without border
double vol= 0.0 , mass = 0.0, ie = 0.0, temp = 0.0;
ops_par_loop(field_summary_kernel, "field_summary_kernel", tea_grid, 2, rangexy_inner,
ops_arg_dat(volume, 1, S2D_00, "double", OPS_READ),
ops_arg_dat(density, 1, S2D_00, "double", OPS_READ),
ops_arg_dat(energy1, 1, S2D_00, "double", OPS_READ),
ops_arg_dat(u, 1, S2D_00, "double", OPS_READ),
ops_arg_reduce(red_vol, 1, "double", OPS_INC),
ops_arg_reduce(red_mass, 1, "double", OPS_INC),
ops_arg_reduce(red_ie, 1, "double", OPS_INC),
ops_arg_reduce(red_temp, 1, "double", OPS_INC));
//printf("mass = %lf\n",mass);
ops_reduction_result(red_vol,&vol);
ops_reduction_result(red_mass,&mass);
ops_reduction_result(red_ie,&ie);
ops_reduction_result(red_temp,&temp);
ops_fprintf(g_out,"\n");
ops_fprintf(g_out,"\n Time %lf\n",clover_time);
ops_fprintf(g_out," %-10s %-10s %-15s %-10s %-s\n",
" Volume"," Mass"," Density"," Internal Energy","Temperature");
ops_fprintf(g_out," step: %3d %-10.3E %-10.3E %-15.3E %-10.3E %-.3E",
step, vol, mass, mass/vol, ie, temp);
if(complete == 1) {
if(test_problem>0) {
if (test_problem == 1)
qa_diff = fabs((100.0 * (temp / 157.55084183279294)) - 100.0);
if (test_problem == 2) // tea_bm_short.in
qa_diff = fabs((100.0 * (temp / 106.27221178646569)) - 100.0);
if (test_problem == 3)
qa_diff = fabs((100.0 * (temp / 99.955877498324000)) - 100.0);
if (test_problem == 4)
qa_diff = fabs((100.0 * (temp / 97.277332050749976)) - 100.0);
if (test_problem == 5)
qa_diff = fabs((100.0 * (temp / 95.462351583362249)) - 100.0);
ops_printf("Test problem %3d is within %-10.7E%% of the expected solution\n",test_problem, qa_diff);
ops_fprintf(g_out,"\nTest problem %3d is within %10.7E%% of the expected solution\n",test_problem, qa_diff);
if(qa_diff < 0.001) {
ops_printf(" This test is considered PASSED\n");
ops_fprintf(g_out," This test is considered PASSED\n");
}
else
{
ops_printf(" This test is considered FAILED\n");
ops_fprintf(g_out," This test is considered FAILED\n");
}
}
}
fflush(g_out);
//ops_exit();//exit for now
//exit(0);
}
int main(int argc, const char **argv) {
c0 = 0.500000000000000;
rc0 = 1.0 / 280.0;
rc1 = 4.0 / 105.0;
rc2 = 1.0 / 5.0;
rc3 = 4.0 / 5.0;
nx0 = 1000;
deltai0 = 0.00100000000000000;
deltat = 0.000400000000000000;
rkold[0] = 1.0 / 4.0;
rkold[1] = 3.0 / 20.0;
rkold[2] = 3.0 / 5.0;
rknew[0] = 2.0 / 3.0;
rknew[1] = 5.0 / 12.0;
rknew[2] = 3.0 / 5.0;
ops_init(argc, argv, 1);
ops_init_backend();
ops_decl_const2("c0", 1, "double", &c0);
ops_decl_const2("rc0", 1, "double", &rc0);
ops_decl_const2("rc1", 1, "double", &rc1);
ops_decl_const2("rc2", 1, "double", &rc2);
ops_decl_const2("rc3", 1, "double", &rc3);
ops_decl_const2("nx0", 1, "int", &nx0);
ops_decl_const2("deltai0", 1, "double", &deltai0);
ops_decl_const2("deltat", 1, "double", &deltat);
ops_block complex_numbers_block;
complex_numbers_block = ops_decl_block(1, "complex_numbers_block");
ops_dat phi;
ops_dat phi_old;
ops_dat wk0;
ops_dat wk1;
int halo_p[] = {4};
int halo_m[] = {-4};
int size[] = {nx0};
int base[] = {0};
double *val = NULL;
phi = ops_decl_dat(complex_numbers_block, 1, size, base, halo_m, halo_p, val,
"double", "phi");
phi_old = ops_decl_dat(complex_numbers_block, 1, size, base, halo_m, halo_p,
val, "double", "phi_old");
wk0 = ops_decl_dat(complex_numbers_block, 1, size, base, halo_m, halo_p, val,
"double", "wk0");
wk1 = ops_decl_dat(complex_numbers_block, 1, size, base, halo_m, halo_p, val,
"double", "wk1");
int stencil1_temp[] = {0};
ops_stencil stencil1 = ops_decl_stencil(1, 1, stencil1_temp, "0");
int stencil0_temp[] = {-4, -3, -2, -1, 1, 2, 3, 4};
ops_stencil stencil0 =
ops_decl_stencil(1, 8, stencil0_temp, "-4,-3,-2,-1,1,2,3,4");
ops_reduction real =
ops_decl_reduction_handle(sizeof(double), "double", "reduction_real");
ops_reduction imaginary = ops_decl_reduction_handle(sizeof(double), "double",
"reduction_imaginary");
ops_halo_group halo_exchange0;
{
int halo_iter[] = {4};
int from_base[] = {0};
int to_base[] = {nx0};
int dir[] = {1};
ops_halo halo0 =
ops_decl_halo(phi, phi, halo_iter, from_base, to_base, dir, dir);
ops_halo grp[] = {halo0};
halo_exchange0 = ops_decl_halo_group(1, grp);
}
ops_halo_group halo_exchange1;
{
int halo_iter[] = {4};
int from_base[] = {nx0 - 4};
int to_base[] = {-4};
int dir[] = {1};
ops_halo halo0 =
ops_decl_halo(phi, phi, halo_iter, from_base, to_base, dir, dir);
ops_halo grp[] = {halo0};
halo_exchange1 = ops_decl_halo_group(1, grp);
}
ops_partition("");
int iter_range5[] = {-4, nx0 + 4};
ops_par_loop_complex_numbers_block0_5_kernel(
"Initialisation", complex_numbers_block, 1, iter_range5,
ops_arg_dat(phi, 1, stencil1, "double", OPS_WRITE), ops_arg_idx());
ops_halo_transfer(halo_exchange0);
ops_halo_transfer(halo_exchange1);
double cpu_start, elapsed_start;
ops_timers(&cpu_start, &elapsed_start);
for (int iteration = 0; iteration < 1; iteration++) {
int iter_range4[] = {-4, nx0 + 4};
ops_par_loop_complex_numbers_block0_4_kernel(
"Save equations", complex_numbers_block, 1, iter_range4,
ops_arg_dat(phi, 1, stencil1, "double", OPS_READ),
ops_arg_dat(phi_old, 1, stencil1, "double", OPS_WRITE));
for (int stage = 0; stage < 3; stage++) {
int iter_range0[] = {0, nx0};
ops_par_loop_complex_numbers_block0_0_kernel(
"D(phi[x0 t] x0)", complex_numbers_block, 1, iter_range0,
ops_arg_dat(phi, 1, stencil0, "double", OPS_READ),
ops_arg_dat(wk0, 1, stencil1, "double", OPS_WRITE));
int iter_range1[] = {0, nx0};
ops_par_loop_complex_numbers_block0_1_kernel(
"Residual of equation", complex_numbers_block, 1, iter_range1,
ops_arg_dat(wk0, 1, stencil1, "double", OPS_READ),
ops_arg_dat(wk1, 1, stencil1, "double", OPS_WRITE));
int iter_range2[] = {-4, nx0 + 4};
ops_par_loop_complex_numbers_block0_2_kernel(
"RK new (subloop) update", complex_numbers_block, 1, iter_range2,
ops_arg_dat(phi_old, 1, stencil1, "double", OPS_READ),
ops_arg_dat(wk1, 1, stencil1, "double", OPS_READ),
ops_arg_dat(phi, 1, stencil1, "double", OPS_WRITE),
ops_arg_gbl(&rknew[stage], 1, "double", OPS_READ));
int iter_range3[] = {-4, nx0 + 4};
ops_par_loop_complex_numbers_block0_3_kernel(
"RK old update", complex_numbers_block, 1, iter_range3,
ops_arg_dat(wk1, 1, stencil1, "double", OPS_READ),
ops_arg_dat(phi_old, 1, stencil1, "double", OPS_RW),
ops_arg_gbl(&rkold[stage], 1, "double", OPS_READ));
ops_halo_transfer(halo_exchange0);
ops_halo_transfer(halo_exchange1);
}
int iter_range0[] = {0, nx0};
ops_par_loop_complex_numbers_block0_cn_kernel(
"Complex numbers", complex_numbers_block, 1, iter_range0,
ops_arg_dat(phi, 1, stencil0, "double", OPS_READ),
ops_arg_reduce(real, 1, "double", OPS_INC),
ops_arg_reduce(imaginary, 1, "double", OPS_INC));
}
double cpu_end, elapsed_end;
ops_timers(&cpu_end, &elapsed_end);
ops_printf("\nTimings are:\n");
ops_printf("-----------------------------------------\n");
ops_printf("Total Wall time %lf\n", elapsed_end - elapsed_start);
ops_fetch_block_hdf5_file(complex_numbers_block, "complex_numbers_2500.h5");
ops_fetch_dat_hdf5_file(phi, "complex_numbers_2500.h5");
ops_exit();
}
void calc_dt(double* local_dt, char* local_control,
double* xl_pos, double* yl_pos, int* jldt, int* kldt)
{
int small;
double jk_control = 1.1;
small = 0;
int dtl_control;
//initialize sizes using global values
int x_min = field.x_min;
int x_max = field.x_max;
int y_min = field.y_min;
int y_max = field.y_max;
int rangexy_inner[] = {x_min,x_max,y_min,y_max}; // inner range without border
ops_par_loop(calc_dt_kernel, "calc_dt_kernel", clover_grid, 2, rangexy_inner,
ops_arg_dat(celldx, 1, S2D_00_P10_STRID2D_X, "double", OPS_READ),
ops_arg_dat(celldy, 1, S2D_00_0P1_STRID2D_Y, "double", OPS_READ),
ops_arg_dat(soundspeed, 1, S2D_00, "double", OPS_READ),
ops_arg_dat(viscosity, 1, S2D_00, "double", OPS_READ),
ops_arg_dat(density0, 1, S2D_00, "double", OPS_READ),
ops_arg_dat(xvel0, 1, S2D_00_P10_0P1_P1P1, "double", OPS_READ),
ops_arg_dat(xarea, 1, S2D_00_P10, "double", OPS_READ),
ops_arg_dat(volume, 1, S2D_00, "double", OPS_READ),
ops_arg_dat(yvel0, 1, S2D_00_P10_0P1_P1P1, "double", OPS_READ),
ops_arg_dat(yarea, 1, S2D_00_0P1, "double", OPS_READ),
ops_arg_dat(work_array1, 1, S2D_00, "double", OPS_WRITE) );
ops_par_loop(calc_dt_kernel_min, "calc_dt_kernel_min", clover_grid, 2, rangexy_inner,
ops_arg_dat(work_array1, 1, S2D_00, "double", OPS_READ),
ops_arg_reduce(red_local_dt, 1, "double", OPS_MIN));
//printf("*local_dt = %lf\n",*local_dt);
//Extract the mimimum timestep information
dtl_control = 10.01 * (jk_control - (int)(jk_control));
jk_control = jk_control - (jk_control - (int)(jk_control));
//*jldt = ((int)jk_control)%x_max;
//*kldt = 1 + (jk_control/x_max);
*jldt = ((int)jk_control)%(x_max-2);
*kldt = 1 + (jk_control/(x_max-2));
int rangexy_getpoint[] = {*jldt-1+2,*jldt+2,*kldt-1+2,*kldt+2}; // get point value //note +2 added due to boundary
//int rangexy_getpointx[] = {*jldt-1+2,*jldt+2,y_min-2,y_max+2}; // get point value //note +2 added due to boundary
//int rangexy_getpointy[] = {x_min-2,x_max+2,*kldt-1+2,*kldt+2}; // get point value //note +2 added due to boundary
ops_par_loop(calc_dt_kernel_get, "calc_dt_kernel_getx", clover_grid, 2, rangexy_getpoint,
ops_arg_dat(cellx, 1, S2D_00_STRID2D_X, "double", OPS_READ),
ops_arg_dat(celly, 1, S2D_00_STRID2D_Y, "double", OPS_READ),
ops_arg_reduce(red_xl_pos, 1, "double", OPS_INC),
ops_arg_reduce(red_yl_pos, 1, "double", OPS_INC));
ops_reduction_result(red_local_dt, local_dt);
ops_reduction_result(red_xl_pos, xl_pos);
ops_reduction_result(red_yl_pos, yl_pos);
*local_dt = MIN(*local_dt, g_big);
if(*local_dt < dtmin) small = 1;
if(small != 0) {
ops_printf("Timestep information:\n");
ops_printf("j, k : %d, %d\n",*jldt,*kldt);
ops_printf("x, y : %lf, %lf\n",*xl_pos,*xl_pos);
ops_printf("timestep : %lf\n",*local_dt);
double output[12] = {0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0};
ops_par_loop(calc_dt_kernel_print, "calc_dt_kernel_print", clover_grid, 2,rangexy_getpoint,
ops_arg_dat(xvel0, 1, S2D_10_M10_01_0M1, "double", OPS_READ),
ops_arg_dat(yvel0, 1, S2D_10_M10_01_0M1, "double", OPS_READ),
ops_arg_dat(density0, 1, S2D_00, "double", OPS_READ),
ops_arg_dat(energy0, 1, S2D_00, "double", OPS_READ),
ops_arg_dat(pressure, 1, S2D_00, "double", OPS_READ),
ops_arg_dat(soundspeed, 1, S2D_00, "double", OPS_READ),
ops_arg_reduce(red_output, 12, "double", OPS_INC));
ops_reduction_result(red_output, output);
ops_printf("Cell velocities:\n");
ops_printf("%E, %E \n",output[0],output[1]); //xvel0(jldt ,kldt ),yvel0(jldt ,kldt )
ops_printf("%E, %E \n",output[2],output[3]); //xvel0(jldt+1,kldt ),yvel0(jldt+1,kldt )
ops_printf("%E, %E \n",output[4],output[5]); //xvel0(jldt+1,kldt+1),yvel0(jldt+1,kldt+1)
ops_printf("%E, %E \n",output[6],output[7]); //xvel0(jldt ,kldt+1),yvel0(jldt ,kldt+1)
ops_printf("density, energy, pressure, soundspeed = %lf, %lf, %lf, %lf \n",
output[8], output[9], output[10], output[11]);
}
if(dtl_control == 1) sprintf(local_control, "sound");
if(dtl_control == 2) sprintf(local_control, "xvel");
if(dtl_control == 3) sprintf(local_control, "yvel");
if(dtl_control == 4) sprintf(local_control, "div");
}
void read_input()
{
//some defailt values before read input
test_problem = 0;
state_max = 0;
number_of_states = 0;
//grid = (grid_type ) xmalloc(sizeof(grid_type_core));
grid.xmin = 0;
grid.ymin = 0;
grid.zmin = 0;
grid.xmax = 100;
grid.ymax = 100;
grid.zmax = 100;
grid.x_cells = 10;
grid.y_cells = 10;
grid.z_cells = 10;
end_time = 10.0;
end_step = g_ibig;
complete = FALSE;
visit_frequency=10;
summary_frequency=10;
dtinit = 0.1;
dtmax = 1.0;
dtmin = 0.0000001;
dtrise = 1.05;
dtc_safe = 0.75;
dtu_safe = 0.5;
dtv_safe = 0.5;
dtw_safe = 0.5;
dtdiv_safe = 0.5;
use_vector_loops = TRUE;
//
//need to read in the following through I/O .. hard coded below
//
ops_fprintf(g_out," Reading input file\n");
#define LINESZ 1024
char buff[LINESZ];
FILE *fin = fopen ("clover.in", "r");
if (fin != NULL) {
while (fgets (buff, LINESZ, fin)) {
char* token = strtok(buff, " =");
while (token) {
if(strcmp(token,"*clover\n") != 0 && strcmp(token,"*endclover\n") != 0 ) {
//printf("token: %s ", token);
if(strcmp(token,"initial_timestep") == 0) {
token = strtok(NULL, " =");
dtinit = atof(token);
ops_fprintf(g_out," %20s: %e\n", "initial_timestep",dtinit);
}
else if(strcmp(token,"max_timestep") == 0) {
token = strtok(NULL, " =");
dtmax = atof(token);
ops_fprintf(g_out," %20s: %e\n", "max_timestep",dtmax);
}
else if(strcmp(token,"timestep_rise") == 0) {
token = strtok(NULL, " =");
dtrise = atof(token);
ops_fprintf(g_out," %20s: %e\n", "timestep_rise",dtrise);
}
else if(strcmp(token,"end_time") == 0) {
token = strtok(NULL, " =");
end_time = atof(token);
ops_fprintf(g_out," %20s: %e\n", "end_time",end_time);
}
else if(strcmp(token,"end_step") == 0) {
token = strtok(NULL, " =");
end_step = atof(token);
ops_fprintf(g_out," %20s: %e\n", "end_step",end_step);
}
else if(strcmp(token,"xmin") == 0) {
token = strtok(NULL, " =");
grid.xmin = atof(token);
ops_fprintf(g_out," %20s: %e\n", "xmin",grid.xmin);
}
else if(strcmp(token,"xmax") == 0) {
token = strtok(NULL, " =");
grid.xmax = atof(token);
ops_fprintf(g_out," %20s: %e\n", "xmax",grid.xmax);
}
else if(strcmp(token,"ymin") == 0) {
token = strtok(NULL, " =");
grid.ymin = atof(token);
ops_fprintf(g_out," %20s: %e\n", "ymin",grid.ymin);
}
else if(strcmp(token,"ymax") == 0) {
token = strtok(NULL, " =");
grid.ymax = atof(token);
ops_fprintf(g_out," %20s: %e\n", "ymax",grid.ymax);
}
else if(strcmp(token,"zmin") == 0) {
token = strtok(NULL, " =");
grid.zmin = atof(token);
ops_fprintf(g_out," %20s: %e\n", "zmin",grid.zmin);
}
else if(strcmp(token,"zmax") == 0) {
token = strtok(NULL, " =");
grid.zmax = atof(token);
ops_fprintf(g_out," %20s: %e\n", "zmax",grid.zmax);
}
else if(strcmp(token,"x_cells") == 0) {
token = strtok(NULL, " =");
grid.x_cells = atof(token);
ops_fprintf(g_out," %20s: %d\n", "x_cells",grid.x_cells);
}
else if(strcmp(token,"y_cells") == 0) {
token = strtok(NULL, " =");
grid.y_cells = atof(token);
ops_fprintf(g_out," %20s: %d\n", "y_cells",grid.y_cells);
}
else if(strcmp(token,"z_cells") == 0) {
token = strtok(NULL, " =");
grid.z_cells = atof(token);
ops_fprintf(g_out," %20s: %d\n", "z_cells",grid.z_cells);
}
else if(strcmp(token,"visit_frequency") == 0) {
token = strtok(NULL, " =");
visit_frequency = atoi(token);
ops_fprintf(g_out," %20s: %d\n", "visit_frequency",visit_frequency);
}
else if(strcmp(token,"summary_frequency") == 0) {
token = strtok(NULL, " =");
summary_frequency = atoi(token);
ops_fprintf(g_out," %20s: %d\n", "summary_frequency",summary_frequency);
}
else if(strcmp(token,"test_problem") == 0) {
token = strtok(NULL, " =");
test_problem = atoi(token);
ops_fprintf(g_out," %20s: %d\n", "test_problem",test_problem);
}
else if(strcmp(token,"profiler_on") == 0) {
token = strtok(NULL, " =");
profiler_on = atoi(token);
ops_fprintf(g_out," %20s: %d\n", "profiler_on",profiler_on);
}
else if(strcmp(token,"state") == 0) {
ops_fprintf(g_out,"\n");
ops_fprintf(g_out," Reading specification for state %d\n",number_of_states+1);
ops_fprintf(g_out,"\n");
token = strtok(NULL, " =");
states = (state_type *) xrealloc(states, sizeof(state_type) * (number_of_states+1));
states[number_of_states].xvel = 0.0;
states[number_of_states].yvel = 0.0;
states[number_of_states].zvel = 0.0;
token = strtok(NULL, " =");
while(token) {
if(strcmp(token,"xvel") == 0) {
token = strtok(NULL, " =");
states[number_of_states].xvel = atof(token);
ops_fprintf(g_out,"xvel: %e\n", states[number_of_states].xvel);
}
if(strcmp(token,"yvel") == 0) {
token = strtok(NULL, " =");
states[number_of_states].yvel = atof(token);
ops_fprintf(g_out,"yvel: %e\n", states[number_of_states].yvel);
}
if(strcmp(token,"zvel") == 0) {
token = strtok(NULL, " =");
states[number_of_states].zvel = atof(token);
ops_fprintf(g_out,"zvel: %e\n", states[number_of_states].zvel);
}
if(strcmp(token,"xmin") == 0) {
token = strtok(NULL, " =");
states[number_of_states].xmin = atof(token);
ops_fprintf(g_out," %20s: %e\n","state xmin",states[number_of_states].xmin);
}
if(strcmp(token,"xmax") == 0) {
token = strtok(NULL, " =");
states[number_of_states].xmax = atof(token);
ops_fprintf(g_out," %20s: %e\n","state xmax",states[number_of_states].xmax);
}
if(strcmp(token,"ymin") == 0) {
token = strtok(NULL, " =");
states[number_of_states].ymin = atof(token);
ops_fprintf(g_out," %20s: %e\n","state ymin",states[number_of_states].ymin);
}
if(strcmp(token,"ymax") == 0) {
token = strtok(NULL, " =");
states[number_of_states].ymax = atof(token);
ops_fprintf(g_out," %20s: %e\n","state ymax",states[number_of_states].ymax);
}
if(strcmp(token,"zmin") == 0) {
token = strtok(NULL, " =");
states[number_of_states].zmin = atof(token);
ops_fprintf(g_out," %20s: %e\n","state zmin",states[number_of_states].zmin);
}
if(strcmp(token,"zmax") == 0) {
token = strtok(NULL, " =");
states[number_of_states].zmax = atof(token);
ops_fprintf(g_out," %20s: %e\n","state zmax",states[number_of_states].zmax);
}
if(strcmp(token,"density") == 0) {
token = strtok(NULL, " =");
states[number_of_states].density = atof(token);
ops_fprintf(g_out," %20s: %e\n", "state density",states[number_of_states].density);
}
if(strcmp(token,"energy") == 0) {
token = strtok(NULL, " =");
states[number_of_states].energy = atof(token);
ops_fprintf(g_out," %20s: %e\n", "state energy",states[number_of_states].energy);
}
if(strcmp(token,"geometry") == 0) {
token = strtok(NULL, " =");
if(strcmp(token,"cuboid") == 0) {
states[number_of_states].geometry = g_cube;
ops_fprintf(g_out," %20s: %s\n","state geometry","cuboid");
}
else if(strcmp(token,"sphere") == 0) {
states[number_of_states].geometry = g_sphe;
ops_fprintf(g_out," %20s: %s\n","state geometry","sphere");
}
else if(strcmp(token,"point") == 0) {
states[number_of_states].geometry = g_point;
ops_fprintf(g_out," %20s: %s\n","state geometry","point");
}
}
token = strtok(NULL, " =");
}
number_of_states++;
ops_fprintf(g_out,"\n");
}
}
token = strtok(NULL, " =");
}
}
fclose (fin);
}
if(number_of_states == 0) {
ops_printf("read_input, No states defined.\n");
exit(-1);
}
ops_fprintf(g_out,"\n");
ops_fprintf(g_out," Input read finished\n");
ops_fprintf(g_out,"\n");
//field = (field_type ) xmalloc(sizeof(field_type_core));
field.x_min = 0 +2; //+2 to account for the boundary
field.y_min = 0 +2; //+2 to account for the boundary
field.z_min = 0 +2; //+2 to account for the boundary
field.x_max = grid.x_cells +2; //+2 to account for the boundary
field.y_max = grid.y_cells +2; //+2 to account for the boundary
field.z_max = grid.z_cells +2; //+2 to account for the boundary
field.left = 0;
field.bottom = 0;
field.back = 0;
float dx= (grid.xmax-grid.xmin)/(float)(grid.x_cells);
float dy= (grid.ymax-grid.ymin)/(float)(grid.y_cells);
float dz= (grid.zmax-grid.zmin)/(float)(grid.z_cells);
for(int i = 0; i < number_of_states; i++)
{
states[i].xmin = states[i].xmin + (dx/100.00);
states[i].ymin = states[i].ymin + (dy/100.00);
states[i].zmin = states[i].zmin + (dz/100.00);
states[i].xmax = states[i].xmax - (dx/100.00);
states[i].ymax = states[i].ymax - (dy/100.00);
states[i].zmax = states[i].zmax - (dz/100.00);
}
}
int main(int argc, char **argv)
{
/**-------------------------- Initialisation --------------------------**/
// OPS initialisation
ops_init(argc,argv,6);
int logical_size_x = 200;
int logical_size_y = 200;
int ngrid_x = 1;
int ngrid_y = 1;
int n_iter = 10000;
dx = 0.01;
dy = 0.01;
ops_decl_const("dx",1,"double",&dx);
ops_decl_const("dy",1,"double",&dy);
//declare blocks
ops_block *blocks = (ops_block *)malloc(ngrid_x*ngrid_y*sizeof(ops_block*));
char buf[50];
for (int j = 0; j < ngrid_y; j++) {
for (int i = 0; i < ngrid_x; i++) {
sprintf(buf,"block %d,%d",i,j);
blocks[i+ngrid_x*j] = ops_decl_block(2,buf);
}
}
//declare stencils
int s2D_00[] = {0,0};
ops_stencil S2D_00 = ops_decl_stencil( 2, 1, s2D_00, "00");
int s2D_00_P10_M10_0P1_0M1[] = {0,0, 1,0, -1,0, 0,1, 0,-1};
ops_stencil S2D_00_P10_M10_0P1_0M1 = ops_decl_stencil( 2, 5, s2D_00_P10_M10_0P1_0M1, "00:10:-10:01:0-1");
ops_reduction red_err = ops_decl_reduction_handle(sizeof(double), "double", "err");
//declare datasets
int d_p[2] = {1,1}; //max halo depths for the dat in the possitive direction
int d_m[2] = {-1,-1}; //max halo depths for the dat in the negative direction
int base[2] = {0,0};
int uniform_size[2] = {(logical_size_x-1)/ngrid_x+1,(logical_size_y-1)/ngrid_y+1};
double* temp = NULL;
ops_dat *coordx = (ops_dat *)malloc(ngrid_x*ngrid_y*sizeof(ops_dat*));
ops_dat *coordy = (ops_dat *)malloc(ngrid_x*ngrid_y*sizeof(ops_dat*));
ops_dat *u = (ops_dat *)malloc(ngrid_x*ngrid_y*sizeof(ops_dat*));
ops_dat *u2 = (ops_dat *)malloc(ngrid_x*ngrid_y*sizeof(ops_dat*));
ops_dat *f = (ops_dat *)malloc(ngrid_x*ngrid_y*sizeof(ops_dat*));
ops_dat *ref = (ops_dat *)malloc(ngrid_x*ngrid_y*sizeof(ops_dat*));
int *sizes = (int*)malloc(2*ngrid_x*ngrid_y*sizeof(int));
int *disps = (int*)malloc(2*ngrid_x*ngrid_y*sizeof(int));
for (int j = 0; j < ngrid_y; j++) {
for (int i = 0; i < ngrid_x; i++) {
int size[2] = {uniform_size[0], uniform_size[1]};
if ((i+1)*size[0]>logical_size_x) size[0] = logical_size_x - i*size[0];
if ((j+1)*size[1]>logical_size_y) size[1] = logical_size_y - j*size[1];
sprintf(buf,"coordx %d,%d",i,j);
coordx[i+ngrid_x*j] = ops_decl_dat(blocks[i+ngrid_x*j], 1, size, base, d_m, d_p, temp, "double", buf);
sprintf(buf,"coordy %d,%d",i,j);
coordy[i+ngrid_x*j] = ops_decl_dat(blocks[i+ngrid_x*j], 1, size, base, d_m, d_p, temp, "double", buf);
sprintf(buf,"u %d,%d",i,j);
u[i+ngrid_x*j] = ops_decl_dat(blocks[i+ngrid_x*j], 1, size, base, d_m, d_p, temp, "double", buf);
sprintf(buf,"u2 %d,%d",i,j);
u2[i+ngrid_x*j] = ops_decl_dat(blocks[i+ngrid_x*j], 1, size, base, d_m, d_p, temp, "double", buf);
sprintf(buf,"f %d,%d",i,j);
f[i+ngrid_x*j] = ops_decl_dat(blocks[i+ngrid_x*j], 1, size, base, d_m, d_p, temp, "double", buf);
sprintf(buf,"ref %d,%d",i,j);
ref[i+ngrid_x*j] = ops_decl_dat(blocks[i+ngrid_x*j], 1, size, base, d_m, d_p, temp, "double", buf);
sizes[2*(i+ngrid_x*j)] = size[0];
sizes[2*(i+ngrid_x*j)+1] = size[1];
disps[2*(i+ngrid_x*j)] = i*uniform_size[0];
disps[2*(i+ngrid_x*j)+1] = j*uniform_size[1];
}
}
ops_halo *halos = (ops_halo *)malloc(2*(ngrid_x*(ngrid_y-1)+(ngrid_x-1)*ngrid_y)*sizeof(ops_halo *));
int off = 0;
for (int j = 0; j < ngrid_y; j++) {
for (int i = 0; i < ngrid_x; i++) {
if (i > 0) {
int halo_iter[] = {1,sizes[2*(i+ngrid_x*j)+1]};
int base_from[] = {sizes[2*(i-1+ngrid_x*j)]-1,0};
int base_to[] = {-1,0};
int dir[] = {1,2};
halos[off++] = ops_decl_halo(u[i-1+ngrid_x*j], u[i+ngrid_x*j], halo_iter, base_from, base_to, dir, dir);
base_from[0] = 0; base_to[0] = sizes[2*(i+ngrid_x*j)];
halos[off++] = ops_decl_halo(u[i+ngrid_x*j], u[i-1+ngrid_x*j], halo_iter, base_from, base_to, dir, dir);
}
if (j > 0) {
int halo_iter[] = {sizes[2*(i+ngrid_x*j)],1};
int base_from[] = {0,sizes[2*(i+ngrid_x*(j-1))+1]-1};
int base_to[] = {0,-1};
int dir[] = {1,2};
halos[off++] = ops_decl_halo(u[i+ngrid_x*(j-1)], u[i+ngrid_x*j], halo_iter, base_from, base_to, dir, dir);
base_from[1] = 0; base_to[1] = sizes[2*(i+ngrid_x*j)+1];
halos[off++] = ops_decl_halo(u[i+ngrid_x*j], u[i+ngrid_x*(j-1)], halo_iter, base_from, base_to, dir, dir);
}
}
}
if (off != 2*(ngrid_x*(ngrid_y-1)+(ngrid_x-1)*ngrid_y)) printf("Something is not right\n");
ops_halo_group u_halos = ops_decl_halo_group(off,halos);
ops_partition("");
ops_checkpointing_init("check.h5", 5.0);
/**-------------------------- Computations --------------------------**/
double ct0, ct1, et0, et1;
ops_timers_core(&ct0, &et0);
//populate forcing, reference solution and boundary conditions
for (int j = 0; j < ngrid_y; j++) {
for (int i = 0; i < ngrid_x; i++) {
int iter_range[] = {-1,sizes[2*(i+ngrid_x*j)]+1,-1,sizes[2*(i+ngrid_x*j)+1]+1};
ops_par_loop(poisson_kernel_populate, "poisson_kernel_populate", blocks[i+ngrid_x*j], 2, iter_range,
ops_arg_gbl(&disps[2*(i+ngrid_x*j)], 1, "int", OPS_READ),
ops_arg_gbl(&disps[2*(i+ngrid_x*j)+1], 1, "int", OPS_READ),
ops_arg_idx(),
ops_arg_dat(u[i+ngrid_x*j], S2D_00, "double", OPS_WRITE),
ops_arg_dat(f[i+ngrid_x*j], S2D_00, "double", OPS_WRITE),
ops_arg_dat(ref[i+ngrid_x*j], S2D_00, "double", OPS_WRITE));
}
}
//initial guess 0
for (int j = 0; j < ngrid_y; j++) {
for (int i = 0; i < ngrid_x; i++) {
int iter_range[] = {0,sizes[2*(i+ngrid_x*j)],0,sizes[2*(i+ngrid_x*j)+1]};
ops_par_loop(poisson_kernel_initialguess, "poisson_kernel_initialguess", blocks[i+ngrid_x*j], 2, iter_range,
ops_arg_dat(u[i+ngrid_x*j], S2D_00, "double", OPS_WRITE));
}
}
for (int iter = 0; iter < n_iter; iter++) {
ops_halo_transfer(u_halos);
for (int j = 0; j < ngrid_y; j++) {
for (int i = 0; i < ngrid_x; i++) {
int iter_range[] = {0,sizes[2*(i+ngrid_x*j)],0,sizes[2*(i+ngrid_x*j)+1]};
ops_par_loop(poisson_kernel_stencil, "poisson_kernel_stencil", blocks[i+ngrid_x*j], 2, iter_range,
ops_arg_dat(u[i+ngrid_x*j], S2D_00_P10_M10_0P1_0M1, "double", OPS_READ),
ops_arg_dat(f[i+ngrid_x*j], S2D_00, "double", OPS_READ),
ops_arg_dat(u2[i+ngrid_x*j], S2D_00, "double", OPS_WRITE));
}
}
for (int j = 0; j < ngrid_y; j++) {
for (int i = 0; i < ngrid_x; i++) {
int iter_range[] = {0,sizes[2*(i+ngrid_x*j)],0,sizes[2*(i+ngrid_x*j)+1]};
ops_par_loop(poisson_kernel_update, "poisson_kernel_update", blocks[i+ngrid_x*j], 2, iter_range,
ops_arg_dat(u2[i+ngrid_x*j], S2D_00, "double", OPS_READ),
ops_arg_dat(u[i+ngrid_x*j] , S2D_00, "double", OPS_WRITE));
}
}
}
double err = 0.0;
for (int j = 0; j < ngrid_y; j++) {
for (int i = 0; i < ngrid_x; i++) {
int iter_range[] = {0,sizes[2*(i+ngrid_x*j)],0,sizes[2*(i+ngrid_x*j)+1]};
ops_par_loop(poisson_kernel_error, "poisson_kernel_error", blocks[i+ngrid_x*j], 2, iter_range,
ops_arg_dat(u[i+ngrid_x*j], S2D_00, "double", OPS_READ),
ops_arg_dat(ref[i+ngrid_x*j] , S2D_00, "double", OPS_READ),
ops_arg_reduce(red_err, 1, "double", OPS_INC));
}
}
ops_reduction_result(red_err,&err);
ops_printf("Total error: %g\n",err);
ops_timers_core(&ct1, &et1);
ops_timing_output();
ops_printf("\nTotal Wall time %lf\n",et1-et0);
ops_exit();
}
void field_summary() {
double qa_diff;
int x_min = field.x_min;
int x_max = field.x_max;
int y_min = field.y_min;
int y_max = field.y_max;
int z_min = field.z_min;
int z_max = field.z_max;
int rangexyz_inner[] = {x_min, x_max, y_min, y_max, z_min, z_max};
ideal_gas(FALSE);
double vol = 0.0, mass = 0.0, ie = 0.0, ke = 0.0, press = 0.0;
ops_par_loop_field_summary_kernel(
"field_summary_kernel", clover_grid, 3, rangexyz_inner,
ops_arg_dat(volume, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(density0, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(energy0, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(pressure, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(xvel0, 1, S3D_000_fP1P1P1, "double", OPS_READ),
ops_arg_dat(yvel0, 1, S3D_000_fP1P1P1, "double", OPS_READ),
ops_arg_dat(zvel0, 1, S3D_000_fP1P1P1, "double", OPS_READ),
ops_arg_reduce(red_vol, 1, "double", OPS_INC),
ops_arg_reduce(red_mass, 1, "double", OPS_INC),
ops_arg_reduce(red_ie, 1, "double", OPS_INC),
ops_arg_reduce(red_ke, 1, "double", OPS_INC),
ops_arg_reduce(red_press, 1, "double", OPS_INC));
ops_reduction_result(red_vol, &vol);
ops_reduction_result(red_mass, &mass);
ops_reduction_result(red_ie, &ie);
ops_reduction_result(red_ke, &ke);
ops_reduction_result(red_press, &press);
ops_fprintf(g_out, "\n");
ops_fprintf(g_out, "\n Time %lf\n", clover_time);
ops_fprintf(g_out,
" %-10s %-10s %-10s %-10s %-15s %-15s %-s\n",
" Volume", " Mass", " Density", " Pressure", " Internal Energy",
"Kinetic Energy", "Total Energy");
ops_fprintf(g_out, " step: %3d %-10.3E %-10.3E %-10.3E %-10.3E "
"%-15.3E %-15.3E %-.3E",
step, vol, mass, mass / vol, press / vol, ie, ke, ie + ke);
if (complete == TRUE && test_problem) {
qa_diff = DBL_MAX;
if (test_problem == 1)
qa_diff = fabs((100.0 * (ke / 3.64560737191257)) - 100.0);
if (test_problem == 2)
qa_diff = fabs((100.0 * (ke / 20.0546870878964)) - 100.0);
if (test_problem == 3)
qa_diff = fabs((100.0 * (ke / 0.37517221925665)) - 100.0);
if (test_problem == 4)
qa_diff = fabs((100.0 * (ke / 17.9845165368889)) - 100.0);
if (test_problem == 5)
qa_diff = fabs((100.0 * (ke / 2.05018938455107)) - 100.0);
ops_printf(
"\n\nTest problem %d is within %3.15E %% of the expected solution\n",
test_problem, qa_diff);
ops_fprintf(
g_out,
"\n\nTest problem %d is within %3.15E %% of the expected solution\n",
test_problem, qa_diff);
if (qa_diff < 0.001) {
ops_printf("This test is considered PASSED\n");
ops_fprintf(g_out, "This test is considered PASSED\n");
} else {
ops_printf("This test is considered FAILED\n");
ops_fprintf(g_out, "This test is considered FAILED\n");
}
}
fflush(g_out);
}
