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, 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 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 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);
}
