void reset_field()
{
error_condition = 0; // Not used yet due to issue with OpenA reduction
//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 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}; // inner range without border
ops_par_loop(reset_field_kernel1, "reset_field_kernel1", clover_grid, 3, rangexyz_inner,
ops_arg_dat(density0, S3D_000, "double", OPS_WRITE),
ops_arg_dat(density1, S3D_000, "double", OPS_READ),
ops_arg_dat(energy0, S3D_000, "double", OPS_WRITE),
ops_arg_dat(energy1, S3D_000, "double", OPS_READ));
int rangexyz_inner_plus1xyz[] = {x_min,x_max+1,y_min,y_max+1,z_min,z_max+1}; // inner range without border
ops_par_loop(reset_field_kernel2, "reset_field_kernel2", clover_grid, 3, rangexyz_inner_plus1xyz,
ops_arg_dat(xvel0, S3D_000, "double", OPS_WRITE),
ops_arg_dat(xvel1, S3D_000, "double", OPS_READ),
ops_arg_dat(yvel0, S3D_000, "double", OPS_WRITE),
ops_arg_dat(yvel1, S3D_000, "double", OPS_READ),
ops_arg_dat(zvel0, S3D_000, "double", OPS_WRITE),
ops_arg_dat(zvel1, S3D_000, "double", OPS_READ));
}
void flux_calc()
{
error_condition = 0;
int x_cells = grid.x_cells;
int y_cells = grid.y_cells;
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_plus1x[] = {x_min,x_max+1,y_min,y_max,z_min,z_max};
ops_par_loop_flux_calc_kernelx("flux_calc_kernelx", clover_grid, 3, rangexyz_inner_plus1x,
ops_arg_dat(vol_flux_x, S3D_000, "double", OPS_WRITE),
ops_arg_dat(xarea, S3D_000, "double", OPS_READ),
ops_arg_dat(xvel0, S3D_000_f0P1P1, "double", OPS_READ),
ops_arg_dat(xvel1, S3D_000_f0P1P1, "double", OPS_READ));
int rangexyz_inner_plus1y[] = {x_min,x_max,y_min,y_max+1,z_min,z_max};
ops_par_loop_flux_calc_kernely("flux_calc_kernely", clover_grid, 3, rangexyz_inner_plus1y,
ops_arg_dat(vol_flux_y, S3D_000, "double", OPS_WRITE),
ops_arg_dat(yarea, S3D_000, "double", OPS_READ),
ops_arg_dat(yvel0, S3D_000_fP10P1, "double", OPS_READ),
ops_arg_dat(yvel1, S3D_000_fP10P1, "double", OPS_READ));
int rangexyz_inner_plus1z[] = {x_min,x_max,y_min,y_max,z_min,z_max+1};
ops_par_loop_flux_calc_kernelz("flux_calc_kernelz", clover_grid, 3, rangexyz_inner_plus1z,
ops_arg_dat(vol_flux_z, S3D_000, "double", OPS_WRITE),
ops_arg_dat(zarea, S3D_000, "double", OPS_READ),
ops_arg_dat(zvel0, S3D_000_fP1P10, "double", OPS_READ),
ops_arg_dat(zvel1, S3D_000_fP1P10, "double", OPS_READ));
}
void initialise_chunk()
{
int x_cells = grid.x_cells;
int y_cells = grid.y_cells;
int z_cells = grid.z_cells;
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 rangex[] = {x_min-2, x_max+3, y_min-2, y_max+3, z_min-2, z_max+3};
int rangey[] = {x_min-2, x_max+3, y_min-2, y_max+3, z_min-2, z_max+3};
int rangez[] = {x_min-2, x_max+3, y_min-2, y_max+3, z_min-2, z_max+3};
int rangefull[] = {-2, x_cells+8, -2, y_cells+8, -2, z_cells+8};
ops_par_loop_initialise_chunk_kernel_xx("initialise_chunk_kernel_xx", clover_grid, 3, rangefull,
ops_arg_dat(xx, S3D_000_STRID3D_X, "int", OPS_WRITE),
ops_arg_idx());
ops_par_loop_initialise_chunk_kernel_yy("initialise_chunk_kernel_yy", clover_grid, 3, rangefull,
ops_arg_dat(yy, S3D_000_STRID3D_Y, "int", OPS_WRITE),
ops_arg_idx());
ops_par_loop_initialise_chunk_kernel_zz("initialise_chunk_kernel_zz", clover_grid, 3, rangefull,
ops_arg_dat(zz, S3D_000_STRID3D_Z, "int", OPS_WRITE),
ops_arg_idx());
ops_par_loop_initialise_chunk_kernel_x("initialise_chunk_kernel_x", clover_grid, 3, rangex,
ops_arg_dat(vertexx, S3D_000_STRID3D_X, "double", OPS_WRITE),
ops_arg_dat(xx, S3D_000_STRID3D_X, "int", OPS_READ),
ops_arg_dat(vertexdx, S3D_000_STRID3D_X, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_y("initialise_chunk_kernel_y", clover_grid, 3, rangey,
ops_arg_dat(vertexy, S3D_000_STRID3D_Y, "double", OPS_WRITE),
ops_arg_dat(yy, S3D_000_STRID3D_Y, "int", OPS_READ),
ops_arg_dat(vertexdy, S3D_000_STRID3D_Y, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_z("initialise_chunk_kernel_z", clover_grid, 3, rangez,
ops_arg_dat(vertexz, S3D_000_STRID3D_Z, "double", OPS_WRITE),
ops_arg_dat(zz, S3D_000_STRID3D_Z, "int", OPS_READ),
ops_arg_dat(vertexdz, S3D_000_STRID3D_Z, "double", OPS_WRITE));
rangex[0] = x_min-2; rangex[1] = x_max+2;
ops_par_loop_initialise_chunk_kernel_cellx("initialise_chunk_kernel_cellx", clover_grid, 3, rangex,
ops_arg_dat(vertexx, S3D_000_P100_STRID3D_X, "double", OPS_READ),
ops_arg_dat(cellx, S3D_000_STRID3D_X, "double", OPS_WRITE),
ops_arg_dat(celldx, S3D_000_STRID3D_X, "double", OPS_WRITE));
rangey[2] = y_min-2; rangey[3] = y_max+2;
ops_par_loop_initialise_chunk_kernel_celly("initialise_chunk_kernel_celly", clover_grid, 3, rangey,
ops_arg_dat(vertexy, S3D_000_0P10_STRID3D_Y, "double", OPS_READ),
ops_arg_dat(celly, S3D_000_STRID3D_Y, "double", OPS_WRITE),
ops_arg_dat(celldy, S3D_000_STRID3D_Y, "double", OPS_WRITE));
rangez[4] = z_min-2; rangez[5] = z_max+2;
ops_par_loop_initialise_chunk_kernel_cellz("initialise_chunk_kernel_cellz", clover_grid, 3, rangez,
ops_arg_dat(vertexz, S3D_000_00P1_STRID3D_Z, "double", OPS_READ),
ops_arg_dat(cellz, S3D_000_STRID3D_Z, "double", OPS_WRITE),
ops_arg_dat(celldz, S3D_000_STRID3D_Z, "double", OPS_WRITE));
int rangexyz[] = {x_min-2,x_max+2,y_min-2,y_max+2,z_min-2,z_max+2};
ops_par_loop_initialise_chunk_kernel_volume("initialise_chunk_kernel_volume", clover_grid, 3, rangexyz,
ops_arg_dat(volume, S3D_000, "double", OPS_WRITE),
ops_arg_dat(celldy, S3D_000_STRID3D_Y, "double", OPS_READ),
ops_arg_dat(xarea, S3D_000, "double", OPS_WRITE),
ops_arg_dat(celldx, S3D_000_STRID3D_X, "double", OPS_READ),
ops_arg_dat(yarea, S3D_000, "double", OPS_WRITE),
ops_arg_dat(celldz, S3D_000_STRID3D_Z, "double", OPS_READ),
ops_arg_dat(zarea, S3D_000, "double", OPS_WRITE));
}
void advec_mom(int which_vel, int sweep_number, int dir) {
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[] = {x_min - 2, x_max + 2, y_min - 2,
y_max + 2, z_min - 2, z_max + 2};
ops_dat vel1;
if (which_vel == 1) {
vel1 = xvel1;
} else if (which_vel == 2) {
vel1 = yvel1;
} else if (which_vel == 3) {
vel1 = zvel1;
}
if (sweep_number == 1 && dir == 1) {
ops_par_loop_advec_mom_kernel_x1(
"advec_mom_kernel_x1", clover_grid, 3, rangexyz,
ops_arg_dat(work_array6, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(volume, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(vol_flux_x, 1, S3D_000_P100, "double", OPS_READ),
ops_arg_dat(vol_flux_y, 1, S3D_000_0P10, "double", OPS_READ),
ops_arg_dat(vol_flux_z, 1, S3D_000_00P1, "double", OPS_READ));
} else if (sweep_number == 1 && dir == 3) {
ops_par_loop_advec_mom_kernel_z1(
"advec_mom_kernel_z1", clover_grid, 3, rangexyz,
ops_arg_dat(work_array6, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(volume, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(vol_flux_x, 1, S3D_000_P100, "double", OPS_READ),
ops_arg_dat(vol_flux_y, 1, S3D_000_0P10, "double", OPS_READ),
ops_arg_dat(vol_flux_z, 1, S3D_000_00P1, "double", OPS_READ));
} else if (sweep_number == 2 && advect_x) {
ops_par_loop_advec_mom_kernel_x2(
"advec_mom_kernel_x2", clover_grid, 3, rangexyz,
ops_arg_dat(work_array6, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(volume, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(vol_flux_y, 1, S3D_000_0P10, "double", OPS_READ),
ops_arg_dat(vol_flux_z, 1, S3D_000_00P1, "double", OPS_READ));
} else if (sweep_number == 2 && !advect_x) {
ops_par_loop_advec_mom_kernel_y2(
"advec_mom_kernel_y2", clover_grid, 3, rangexyz,
ops_arg_dat(work_array6, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(volume, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(vol_flux_x, 1, S3D_000_P100, "double", OPS_READ),
ops_arg_dat(vol_flux_y, 1, S3D_000_0P10, "double", OPS_READ));
} else if (sweep_number == 3 && dir == 1) {
ops_par_loop_advec_mom_kernel_x3(
"advec_mom_kernel_x3", clover_grid, 3, rangexyz,
ops_arg_dat(work_array6, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(volume, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(vol_flux_x, 1, S3D_000_P100, "double", OPS_READ));
} else if (sweep_number == 3 && dir == 3) {
ops_par_loop_advec_mom_kernel_z3(
"advec_mom_kernel_z3", clover_grid, 3, rangexyz,
ops_arg_dat(work_array6, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(volume, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(vol_flux_z, 1, S3D_000_00P1, "double", OPS_READ));
}
if (dir == 1) {
if (which_vel == 1) {
int range_fullx_party_partz_1[] = {x_min - 2, x_max + 2, y_min,
y_max + 1, z_min, z_max + 1};
ops_par_loop_advec_mom_kernel_mass_flux_x(
"advec_mom_kernel_mass_flux_x", clover_grid, 3,
range_fullx_party_partz_1,
ops_arg_dat(work_array1, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(mass_flux_x, 1, S3D_000_fP1M1M1, "double", OPS_READ));
int range_partx_party_partz_1[] = {x_min - 1, x_max + 2, y_min,
y_max + 1, z_min, z_max + 1};
ops_par_loop_advec_mom_kernel_post_pre_advec_x(
"advec_mom_kernel_post_pre_advec_x", clover_grid, 3,
range_partx_party_partz_1,
ops_arg_dat(work_array2, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, 1, S3D_000_fM1M1M1, "double", OPS_READ),
ops_arg_dat(density1, 1, S3D_000_fM1M1M1, "double", OPS_READ),
ops_arg_dat(work_array3, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array1, 1, S3D_000_M100, "double", OPS_READ));
}
int range_innder_plus1xyz_minus1x[] = {x_min - 1, x_max + 1, y_min,
y_max + 1, z_min, z_max + 1};
ops_par_loop_advec_mom_kernel1_x_nonvector(
"advec_mom_kernel1_x", clover_grid, 3, range_innder_plus1xyz_minus1x,
ops_arg_dat(work_array1, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array3, 1, S3D_000_P100, "double", OPS_READ),
ops_arg_dat(work_array5, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(celldx, 1, S3D_000_P100_M100_STRID3D_X, "double", OPS_READ),
ops_arg_dat(vel1, 1, S3D_000_P100_P200_M100, "double", OPS_READ));
int range_partx_party_partz_2[] = {x_min, x_max + 1, y_min,
y_max + 1, z_min, z_max + 1};
ops_par_loop_advec_mom_kernel2_x(
"advec_mom_kernel2_x", clover_grid, 3, range_partx_party_partz_2,
ops_arg_dat(vel1, 1, S3D_000, "double", OPS_RW),
ops_arg_dat(work_array2, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array3, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array5, 1, S3D_000_M100, "double", OPS_READ));
} else if (dir == 2) {
if (which_vel == 1) {
int range_fully_partx_partz_1[] = {x_min, x_max + 1, y_min - 2,
y_max + 2, z_min, z_max + 1};
ops_par_loop_advec_mom_kernel_mass_flux_y(
"advec_mom_kernel_mass_flux_y", clover_grid, 3,
range_fully_partx_partz_1,
ops_arg_dat(work_array1, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(mass_flux_y, 1, S3D_000_fM1P1M1, "double", OPS_READ));
int range_party_partx_partz_1[] = {x_min, x_max + 1, y_min - 1,
y_max + 2, z_min, z_max + 1};
ops_par_loop_advec_mom_kernel_post_pre_advec_y(
"advec_mom_kernel_post_pre_advec_y", clover_grid, 3,
range_party_partx_partz_1,
ops_arg_dat(work_array2, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, 1, S3D_000_fM1M1M1, "double", OPS_READ),
ops_arg_dat(density1, 1, S3D_000_fM1M1M1, "double", OPS_READ),
ops_arg_dat(work_array3, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array1, 1, S3D_000_0M10, "double", OPS_READ));
}
int range_plus1xyz_minus1y[] = {x_min, x_max + 1, y_min - 1,
y_max + 1, z_min, z_max + 1};
ops_par_loop_advec_mom_kernel1_y_nonvector(
"advec_mom_kernel1_y", clover_grid, 3, range_plus1xyz_minus1y,
ops_arg_dat(work_array1, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array3, 1, S3D_000_0P10, "double", OPS_READ),
ops_arg_dat(work_array5, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(celldy, 1, S3D_000_0P10_0M10_STRID3D_Y, "double", OPS_READ),
ops_arg_dat(vel1, 1, S3D_000_0P10_0P20_0M10, "double", OPS_READ));
int range_partx_party_partz_2[] = {x_min, x_max + 1, y_min,
y_max + 1, z_min, z_max + 1};
ops_par_loop_advec_mom_kernel2_y(
"advec_mom_kernel2_y", clover_grid, 3, range_partx_party_partz_2,
ops_arg_dat(vel1, 1, S3D_000, "double", OPS_RW),
ops_arg_dat(work_array2, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array3, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array5, 1, S3D_000_0M10, "double", OPS_READ));
} else if (dir == 3) {
if (which_vel == 1) {
int range_fullz_partx_party_1[] = {x_min, x_max + 1, y_min,
y_max + 1, z_min - 2, z_max + 2};
ops_par_loop_advec_mom_kernel_mass_flux_z(
"advec_mom_kernel_mass_flux_z", clover_grid, 3,
range_fullz_partx_party_1,
ops_arg_dat(work_array1, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(mass_flux_z, 1, S3D_000_fM1M1P1, "double", OPS_READ));
int range_party_partx_partz_1[] = {x_min, x_max + 1, y_min,
y_max + 1, z_min - 1, z_max + 2};
ops_par_loop_advec_mom_kernel_post_pre_advec_z(
"advec_mom_kernel_post_pre_advec_z", clover_grid, 3,
range_party_partx_partz_1,
ops_arg_dat(work_array2, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, 1, S3D_000_fM1M1M1, "double", OPS_READ),
ops_arg_dat(density1, 1, S3D_000_fM1M1M1, "double", OPS_READ),
ops_arg_dat(work_array3, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array1, 1, S3D_000_00M1, "double", OPS_READ));
}
int range_plus1xyz_minus1z[] = {x_min, x_max + 1, y_min,
y_max + 1, z_min - 1, z_max + 1};
ops_par_loop_advec_mom_kernel1_z_nonvector(
"advec_mom_kernel1_z", clover_grid, 3, range_plus1xyz_minus1z,
ops_arg_dat(work_array1, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array3, 1, S3D_000_00P1, "double", OPS_READ),
ops_arg_dat(work_array5, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(celldz, 1, S3D_000_00P1_00M1_STRID3D_Z, "double", OPS_READ),
ops_arg_dat(vel1, 1, S3D_000_00P1_00P2_00M1, "double", OPS_READ));
int range_partx_party_partz_2[] = {x_min, x_max + 1, y_min,
y_max + 1, z_min, z_max + 1};
ops_par_loop_advec_mom_kernel2_z(
"advec_mom_kernel2_z", clover_grid, 3, range_partx_party_partz_2,
ops_arg_dat(vel1, 1, S3D_000, "double", OPS_RW),
ops_arg_dat(work_array2, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array3, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array5, 1, S3D_000_00M1, "double", OPS_READ));
}
}
void advec_mom(int which_vel, int sweep_number, int dir)
{
//initialize sizes using global values
int x_cells = grid.x_cells;
int y_cells = grid.y_cells;
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[] = {x_min-2,x_max+2,y_min-2,y_max+2,z_min-2,z_max+2}; // full range over grid
int mom_sweep;
ops_dat vel1;
if( which_vel == 1) {
vel1 = xvel1;
}
else if( which_vel == 2) {
vel1 = yvel1;
}
else if( which_vel == 3) {
vel1 = zvel1;
}
if(sweep_number==1 && dir == 1) {
ops_par_loop(advec_mom_kernel_x1, "advec_mom_kernel_x1", clover_grid, 3, rangexyz,
ops_arg_dat(work_array6, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, S3D_000, "double", OPS_WRITE),
ops_arg_dat(volume, S3D_000, "double", OPS_READ),
ops_arg_dat(vol_flux_x, S3D_000_P100, "double", OPS_READ),
ops_arg_dat(vol_flux_y, S3D_000_0P10, "double", OPS_READ),
ops_arg_dat(vol_flux_z, S3D_000_00P1, "double", OPS_READ));
}
else if(sweep_number==1 && dir == 3) {
ops_par_loop(advec_mom_kernel_z1, "advec_mom_kernel_z1", clover_grid, 3, rangexyz,
ops_arg_dat(work_array6, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, S3D_000, "double", OPS_WRITE),
ops_arg_dat(volume, S3D_000, "double", OPS_READ),
ops_arg_dat(vol_flux_x, S3D_000_P100, "double", OPS_READ),
ops_arg_dat(vol_flux_y, S3D_000_0P10, "double", OPS_READ),
ops_arg_dat(vol_flux_z, S3D_000_00P1, "double", OPS_READ));
}
else if (sweep_number==2 && advect_x) {
ops_par_loop(advec_mom_kernel_x2, "advec_mom_kernel_x2", clover_grid, 3, rangexyz,
ops_arg_dat(work_array6, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, S3D_000, "double", OPS_WRITE),
ops_arg_dat(volume, S3D_000, "double", OPS_READ),
ops_arg_dat(vol_flux_y, S3D_000_0P10, "double", OPS_READ),
ops_arg_dat(vol_flux_z, S3D_000_00P1, "double", OPS_READ));
}
else if (sweep_number==2 && !advect_x) {
ops_par_loop(advec_mom_kernel_y2, "advec_mom_kernel_y2", clover_grid, 3, rangexyz,
ops_arg_dat(work_array6, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, S3D_000, "double", OPS_WRITE),
ops_arg_dat(volume, S3D_000, "double", OPS_READ),
ops_arg_dat(vol_flux_x, S3D_000_P100, "double", OPS_READ),
ops_arg_dat(vol_flux_y, S3D_000_0P10, "double", OPS_READ));
}
else if (sweep_number==3 && dir == 1) {
ops_par_loop(advec_mom_kernel_x3, "advec_mom_kernel_x3", clover_grid, 3, rangexyz,
ops_arg_dat(work_array6, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, S3D_000, "double", OPS_WRITE),
ops_arg_dat(volume, S3D_000, "double", OPS_READ),
ops_arg_dat(vol_flux_x, S3D_000_P100, "double", OPS_READ));
}
else if (sweep_number==3 && dir == 3) {
ops_par_loop(advec_mom_kernel_z3, "advec_mom_kernel_z3", clover_grid, 3, rangexyz,
ops_arg_dat(work_array6, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, S3D_000, "double", OPS_WRITE),
ops_arg_dat(volume, S3D_000, "double", OPS_READ),
ops_arg_dat(vol_flux_z, S3D_000_00P1, "double", OPS_READ));
}
int range_partx_party_1[] = {x_min-1,x_max+2,y_min,y_max+1}; // partial x range partial y range
int range_fully_party_1[] = {x_min,x_max+1,y_min-2,y_max+2}; // full y range partial x range
int range_partx_party_2[] = {x_min,x_max+1,y_min-1,y_max+2}; // partial x range partial y range
if (dir == 1) {
if (which_vel == 1) {
//Find staggered mesh mass fluxes, nodal masses and volumes.
int range_fullx_party_partz_1[] = {x_min-2,x_max+2,y_min,y_max+1,z_min,z_max+1}; // full x range partial y,z range
ops_par_loop(advec_mom_kernel_mass_flux_x, "advec_mom_kernel_mass_flux_x", clover_grid, 3, range_fullx_party_partz_1,
ops_arg_dat(work_array1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(mass_flux_x, S3D_000_fP1M1M1, "double", OPS_READ));
//Staggered cell mass post and pre advection
int range_partx_party_partz_1[] = {x_min-1,x_max+2,y_min,y_max+1,z_min,z_max+1}; // partial x,y,z range
ops_par_loop(advec_mom_kernel_post_pre_advec_x, "advec_mom_kernel_post_pre_advec_x", clover_grid, 3, range_partx_party_partz_1,
ops_arg_dat(work_array2/*node_mass_post*/, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7/*post_vol*/, S3D_000_fM1M1M1, "double", OPS_READ),
ops_arg_dat(density1, S3D_000_fM1M1M1, "double", OPS_READ),
ops_arg_dat(work_array3/*node_mass_pre*/, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array1/*node_flux*/, S3D_000_M100, "double", OPS_READ));
}
int range_innder_plus1xyz_minus1x[] = {x_min-1,x_max+1,y_min,y_max+1,z_min,z_max+1}; // partial x range partial y range
ops_par_loop(advec_mom_kernel1_x_nonvector, "advec_mom_kernel1_x", clover_grid, 3, range_innder_plus1xyz_minus1x,
ops_arg_dat(work_array1/*node_flux*/, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array3/*node_mass_pre*/, S3D_000_P100, "double", OPS_READ),
ops_arg_dat(work_array5/*mom_flux*/, S3D_000, "double", OPS_WRITE),
ops_arg_dat(celldx, S3D_000_P100_M100_STRID3D_X, "double", OPS_READ),
ops_arg_dat(vel1, S3D_000_P100_P200_M100, "double", OPS_READ));
int range_partx_party_partz_2[] = {x_min,x_max+1,y_min,y_max+1,z_min,z_max+1};
ops_par_loop(advec_mom_kernel2_x, "advec_mom_kernel2_x", clover_grid, 3, range_partx_party_partz_2,
ops_arg_dat(vel1, S3D_000, "double", OPS_RW),
ops_arg_dat(work_array2/*node_mass_post*/, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array3/*node_mass_pre*/, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array5/*mom_flux*/, S3D_000_M100, "double", OPS_READ)
);
}
else if (dir == 2) {
if (which_vel == 1) {
//Find staggered mesh mass fluxes, nodal masses and volumes.
int range_fully_partx_partz_1[] = {x_min,x_max+1,y_min-2,y_max+2,z_min,z_max+1}; // full x range partial y,z range
ops_par_loop(advec_mom_kernel_mass_flux_y, "advec_mom_kernel_mass_flux_y", clover_grid, 3, range_fully_partx_partz_1,
ops_arg_dat(work_array1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(mass_flux_y, S3D_000_fM1P1M1, "double", OPS_READ));
//Staggered cell mass post and pre advection
int range_party_partx_partz_1[] = {x_min,x_max+1,y_min-1,y_max+2,z_min,z_max+1}; // full x range partial y,z range
ops_par_loop(advec_mom_kernel_post_pre_advec_y, "advec_mom_kernel_post_pre_advec_y", clover_grid, 3, range_party_partx_partz_1,
ops_arg_dat(work_array2/*node_mass_post*/, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, S3D_000_fM1M1M1, "double", OPS_READ),
ops_arg_dat(density1, S3D_000_fM1M1M1, "double", OPS_READ),
ops_arg_dat(work_array3/*node_mass_pre*/, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array1/*node_flux*/, S3D_000_0M10, "double", OPS_READ));
}
int range_plus1xyz_minus1y[] = {x_min,x_max+1,y_min-1,y_max+1,z_min,z_max+1}; // partial x range partial y range
ops_par_loop(advec_mom_kernel1_y_nonvector, "advec_mom_kernel1_y", clover_grid, 3, range_plus1xyz_minus1y,
ops_arg_dat(work_array1/*node_flux*/, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array3/*node_mass_pre*/, S3D_000_0P10, "double", OPS_READ),
ops_arg_dat(work_array5/*mom_flux*/, S3D_000, "double", OPS_WRITE),
ops_arg_dat(celldy, S3D_000_0P10_0M10_STRID3D_Y, "double", OPS_READ),
ops_arg_dat(vel1, S3D_000_0P10_0P20_0M10, "double", OPS_READ));
int range_partx_party_partz_2[] = {x_min,x_max+1,y_min,y_max+1,z_min,z_max+1};
ops_par_loop(advec_mom_kernel2_y, "advec_mom_kernel2_y", clover_grid, 3, range_partx_party_partz_2,
ops_arg_dat(vel1, S3D_000, "double", OPS_RW),
ops_arg_dat(work_array2/*node_mass_post*/, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array3/*node_mass_pre*/, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array5/*mom_flux*/, S3D_000_0M10, "double", OPS_READ));
}
else if (dir == 3) {
if (which_vel == 1) {
//Find staggered mesh mass fluxes, nodal masses and volumes.
int range_fullz_partx_party_1[] = {x_min,x_max+1,y_min,y_max+1,z_min-2,z_max+2}; // full x range partial y,z range
ops_par_loop(advec_mom_kernel_mass_flux_z, "advec_mom_kernel_mass_flux_z", clover_grid, 3, range_fullz_partx_party_1,
ops_arg_dat(work_array1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(mass_flux_z, S3D_000_fM1M1P1, "double", OPS_READ));
//Staggered cell mass post and pre advection
int range_party_partx_partz_1[] = {x_min,x_max+1,y_min,y_max+1,z_min-1,z_max+2}; // full x range partial y,z range
ops_par_loop(advec_mom_kernel_post_pre_advec_z, "advec_mom_kernel_post_pre_advec_z", clover_grid, 3, range_party_partx_partz_1,
ops_arg_dat(work_array2/*node_mass_post*/, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, S3D_000_fM1M1M1, "double", OPS_READ),
ops_arg_dat(density1, S3D_000_fM1M1M1, "double", OPS_READ),
ops_arg_dat(work_array3/*node_mass_pre*/, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array1/*node_flux*/, S3D_000_00M1, "double", OPS_READ));
}
int range_plus1xyz_minus1z[] = {x_min,x_max+1,y_min,y_max+1,z_min-1,z_max+1}; // partial x range partial y range
ops_par_loop(advec_mom_kernel1_z_nonvector, "advec_mom_kernel1_z", clover_grid, 3, range_plus1xyz_minus1z,
ops_arg_dat(work_array1/*node_flux*/, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array3/*node_mass_pre*/, S3D_000_00P1, "double", OPS_READ),
ops_arg_dat(work_array5/*mom_flux*/, S3D_000, "double", OPS_WRITE),
ops_arg_dat(celldz, S3D_000_00P1_00M1_STRID3D_Z, "double", OPS_READ),
ops_arg_dat(vel1, S3D_000_00P1_00P2_00M1, "double", OPS_READ));
int range_partx_party_partz_2[] = {x_min,x_max+1,y_min,y_max+1,z_min,z_max+1}; // full x range partial y range
ops_par_loop(advec_mom_kernel2_z, "advec_mom_kernel2_z", clover_grid, 3, range_partx_party_partz_2,
ops_arg_dat(vel1, S3D_000, "double", OPS_RW),
ops_arg_dat(work_array2/*node_mass_post*/, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array3/*node_mass_pre*/, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array5/*mom_flux*/, S3D_000_00M1, "double", OPS_READ));
}
}
void initialise_chunk() {
int x_cells = grid.x_cells;
int y_cells = grid.y_cells;
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 range[] = {x_min, x_max, y_min, y_max};
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, range,
ops_arg_dat(density, 1, S2D_00, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, range,
ops_arg_dat(energy0, 1, S2D_00, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, range,
ops_arg_dat(energy1, 1, S2D_00, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, range,
ops_arg_dat(u, 1, S2D_00, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, range,
ops_arg_dat(u0, 1, S2D_00, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, range,
ops_arg_dat(vector_r, 1, S2D_00, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, range,
ops_arg_dat(vector_rstore, 1, S2D_00, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, range,
ops_arg_dat(vector_rtemp, 1, S2D_00, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, range,
ops_arg_dat(vector_Mi, 1, S2D_00, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, range,
ops_arg_dat(vector_w, 1, S2D_00, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, range,
ops_arg_dat(vector_z, 1, S2D_00, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, range,
ops_arg_dat(vector_utemp, 1, S2D_00, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, range,
ops_arg_dat(vector_Kx, 1, S2D_00, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, range,
ops_arg_dat(vector_Ky, 1, S2D_00, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, range,
ops_arg_dat(vector_p, 1, S2D_00, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, range,
ops_arg_dat(vector_sd, 1, S2D_00, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, range,
ops_arg_dat(tri_cp, 1, S2D_00, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, range,
ops_arg_dat(tri_bfp, 1, S2D_00, "double", OPS_WRITE));
int rangefull1[] = {-2, x_cells + 2, -2, y_cells + 2};
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, rangefull1,
ops_arg_dat(volume, 1, S2D_00, "double", OPS_WRITE));
int rangefull2[] = {-2, x_cells + 3, -2, y_cells + 2};
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, rangefull2,
ops_arg_dat(xarea, 1, S2D_00, "double", OPS_WRITE));
int rangefull3[] = {-2, x_cells + 2, -2, y_cells + 3};
ops_par_loop_initialise_chunk_kernel_zero(
"initialise_chunk_kernel_zero", tea_grid, 2, rangefull3,
ops_arg_dat(yarea, 1, S2D_00, "double", OPS_WRITE));
int rangex[] = {x_min - 2, x_max + 2, y_min - 2, y_max + 2};
ops_par_loop_initialise_chunk_kernel_zero_x(
"initialise_chunk_kernel_zero_x", tea_grid, 2, rangex,
ops_arg_dat(cellx, 1, S2D_00_STRID2D_X, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero_x(
"initialise_chunk_kernel_zero_x", tea_grid, 2, rangex,
ops_arg_dat(celldx, 1, S2D_00_STRID2D_X, "double", OPS_WRITE));
rangex[1]++;
ops_par_loop_initialise_chunk_kernel_zero_x(
"initialise_chunk_kernel_zero_x", tea_grid, 2, rangex,
ops_arg_dat(vertexx, 1, S2D_00_STRID2D_X, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero_x(
"initialise_chunk_kernel_zero_x", tea_grid, 2, rangex,
ops_arg_dat(vertexdx, 1, S2D_00_STRID2D_X, "double", OPS_WRITE));
int rangey2[] = {x_min - 2, x_max + 2, y_min - 2, y_max + 2};
ops_par_loop_initialise_chunk_kernel_zero_y(
"initialise_chunk_kernel_zero_y", tea_grid, 2, rangey2,
ops_arg_dat(celly, 1, S2D_00_STRID2D_Y, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero_y(
"initialise_chunk_kernel_zero_y", tea_grid, 2, rangey2,
ops_arg_dat(celldy, 1, S2D_00_STRID2D_Y, "double", OPS_WRITE));
rangey2[3]++;
ops_par_loop_initialise_chunk_kernel_zero_y(
"initialise_chunk_kernel_zero_y", tea_grid, 2, rangey2,
ops_arg_dat(vertexy, 1, S2D_00_STRID2D_Y, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_zero_y(
"initialise_chunk_kernel_zero_y", tea_grid, 2, rangey2,
ops_arg_dat(vertexdy, 1, S2D_00_STRID2D_Y, "double", OPS_WRITE));
int rangefull[] = {x_min - 2, x_max + 3, y_min - 2, y_max + 3};
ops_execute();
ops_par_loop_initialise_chunk_kernel_xx(
"initialise_chunk_kernel_xx", tea_grid, 2, rangefull,
ops_arg_dat(xx, 1, S2D_00_STRID2D_X, "int", OPS_WRITE), ops_arg_idx());
ops_par_loop_initialise_chunk_kernel_yy(
"initialise_chunk_kernel_yy", tea_grid, 2, rangefull,
ops_arg_dat(yy, 1, S2D_00_STRID2D_Y, "int", OPS_WRITE), ops_arg_idx());
ops_par_loop_initialise_chunk_kernel_x(
"initialise_chunk_kernel_x", tea_grid, 2, rangex,
ops_arg_dat(vertexx, 1, S2D_00_STRID2D_X, "double", OPS_WRITE),
ops_arg_dat(xx, 1, S2D_00_STRID2D_X, "int", OPS_READ),
ops_arg_dat(vertexdx, 1, S2D_00_STRID2D_X, "double", OPS_WRITE));
ops_par_loop_initialise_chunk_kernel_y(
"initialise_chunk_kernel_y", tea_grid, 2, rangey2,
ops_arg_dat(vertexy, 1, S2D_00_STRID2D_Y, "double", OPS_WRITE),
ops_arg_dat(yy, 1, S2D_00_STRID2D_Y, "int", OPS_READ),
ops_arg_dat(vertexdy, 1, S2D_00_STRID2D_Y, "double", OPS_WRITE));
rangex[0] = x_min - 2;
rangex[1] = x_max + 2;
rangex[2] = y_min - 2;
rangex[3] = y_max + 2;
ops_par_loop_initialise_chunk_kernel_cellx(
"initialise_chunk_kernel_cellx", tea_grid, 2, rangex,
ops_arg_dat(vertexx, 1, S2D_00_P10_STRID2D_X, "double", OPS_READ),
ops_arg_dat(cellx, 1, S2D_00_STRID2D_X, "double", OPS_WRITE),
ops_arg_dat(celldx, 1, S2D_00_STRID2D_X, "double", OPS_WRITE));
int rangey[] = {x_min - 2, x_max + 3, y_min - 2, y_max + 2};
ops_par_loop_initialise_chunk_kernel_celly(
"initialise_chunk_kernel_celly", tea_grid, 2, rangey,
ops_arg_dat(vertexy, 1, S2D_00_0P1_STRID2D_Y, "double", OPS_READ),
ops_arg_dat(celly, 1, S2D_00_STRID2D_Y, "double", OPS_WRITE),
ops_arg_dat(celldy, 1, S2D_00_STRID2D_Y, "double", OPS_WRITE));
int rangexy[] = {x_min - 2, x_max + 2, y_min - 2, y_max + 2};
ops_par_loop_initialise_chunk_kernel_volume(
"initialise_chunk_kernel_volume", tea_grid, 2, rangexy,
ops_arg_dat(volume, 1, S2D_00, "double", OPS_WRITE),
ops_arg_dat(celldy, 1, S2D_00_STRID2D_Y, "double", OPS_READ),
ops_arg_dat(xarea, 1, S2D_00, "double", OPS_WRITE),
ops_arg_dat(celldx, 1, S2D_00_STRID2D_X, "double", OPS_READ),
ops_arg_dat(yarea, 1, S2D_00, "double", OPS_WRITE));
}
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 generate()
{
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[] = {x_min-2,x_max+2,y_min-2,y_max+2,z_min-2,z_max+2};
ops_par_loop_generate_chunk_kernel("generate_chunk_kernel", clover_grid, 3, rangexyz,
ops_arg_dat(vertexx, S3D_000_P100_STRID3D_X, "double", OPS_READ),
ops_arg_dat(vertexy, S3D_000_0P10_STRID3D_Y, "double", OPS_READ),
ops_arg_dat(vertexz, S3D_000_00P1_STRID3D_Z, "double", OPS_READ),
ops_arg_dat(energy0, S3D_000, "double", OPS_WRITE),
ops_arg_dat(density0, S3D_000, "double", OPS_WRITE),
ops_arg_dat(xvel0, S3D_000_fP1P1P1, "double", OPS_WRITE),
ops_arg_dat(yvel0, S3D_000_fP1P1P1, "double", OPS_WRITE),
ops_arg_dat(zvel0, S3D_000_fP1P1P1, "double", OPS_WRITE),
ops_arg_dat(cellx, S3D_000_STRID3D_X, "double", OPS_READ),
ops_arg_dat(celly, S3D_000_STRID3D_Y, "double", OPS_READ),
ops_arg_dat(cellz, S3D_000_STRID3D_Z, "double", OPS_READ));
}
void tea_leaf_cheby_init(
ops_dat u,
ops_dat u0,
ops_dat p,
ops_dat r,
ops_dat Mi,
ops_dat w,
ops_dat z,
ops_dat Kx,
ops_dat Ky,
ops_dat cp,
ops_dat bfp,
double rx, double ry,
double theta, int preconditioner_type)
{
//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[] = {x_min,x_max,y_min,y_max};
ops_par_loop(tea_leaf_cheby_init_kernel, "tea_leaf_cheby_init_kernel", tea_grid, 2, rangexy,
ops_arg_dat(w, 1, S2D_00, "double", OPS_WRITE),
ops_arg_dat(r, 1, S2D_00, "double", OPS_WRITE),
ops_arg_dat(Kx, 1, S2D_00_P10, "double", OPS_READ),
ops_arg_dat(Ky, 1, S2D_00_0P1, "double", OPS_READ),
ops_arg_dat(u, 1, S2D_00_0M1_M10_P10_0P1, "double", OPS_READ),
ops_arg_dat(u0, 1, S2D_00, "double", OPS_READ),
ops_arg_gbl(&rx, 1, "double", OPS_READ),
ops_arg_gbl(&ry, 1, "double", OPS_READ));
if (preconditioner_type != TL_PREC_NONE) {
if (preconditioner_type == TL_PREC_JAC_BLOCK)
tea_block_solve(r, z, cp, bfp, Kx, Ky, rx, ry);
else if (preconditioner_type == TL_PREC_JAC_DIAG)
tea_diag_solve(r, z, Mi, Kx, Ky, rx, ry);
ops_par_loop(tea_leaf_recip3_kernel, "tea_leaf_recip3_kernel", tea_grid, 2, rangexy,
ops_arg_dat(p, 1, S2D_00, "double", OPS_WRITE),
ops_arg_dat(z, 1, S2D_00, "double", OPS_READ),
ops_arg_gbl(&theta, 1, "double", OPS_READ));
} else {
ops_par_loop(tea_leaf_recip3_kernel, "tea_leaf_recip3_kernel", tea_grid, 2, rangexy,
ops_arg_dat(p, 1, S2D_00, "double", OPS_WRITE),
ops_arg_dat(r, 1, S2D_00, "double", OPS_READ),
ops_arg_gbl(&theta, 1, "double", OPS_READ));
}
double one = 1.0;
ops_par_loop(tea_leaf_xpy_kernel, "tea_leaf_xpy_kernel", tea_grid, 2, rangexy,
ops_arg_dat(u, 1, S2D_00, "double", OPS_INC),
ops_arg_dat(p, 1, S2D_00, "double", OPS_READ));
}
void accelerate()
{
error_condition = 0; // Not used yet due to issue with OpenA reduction
//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_plus1[] = {x_min,x_max+1,y_min,y_max+1}; // inner range plus 1
ops_par_loop(accelerate_kernel, "accelerate_kernel", clover_grid, 2, rangexy_inner_plus1,
ops_arg_dat(density0, 1, S2D_00_M10_0M1_M1M1, "double", OPS_READ),
ops_arg_dat(volume, 1, S2D_00_M10_0M1_M1M1, "double", OPS_READ),
ops_arg_dat(work_array1, 1, S2D_00, "double", OPS_WRITE),
ops_arg_dat(xvel0, 1, S2D_00, "double", OPS_READ),
ops_arg_dat(xvel1, 1, S2D_00, "double", OPS_INC),
ops_arg_dat(xarea, 1, S2D_00_0M1, "double", OPS_READ),
ops_arg_dat(pressure, 1, S2D_00_M10_0M1_M1M1, "double", OPS_READ),
ops_arg_dat(yvel0, 1, S2D_00, "double", OPS_READ),
ops_arg_dat(yvel1, 1, S2D_00, "double", OPS_INC),
ops_arg_dat(yarea, 1, S2D_00_M10, "double", OPS_READ),
ops_arg_dat(viscosity, 1, S2D_00_M10_0M1_M1M1, "double", OPS_READ));
}
void tea_leaf_cheby_iterate(
ops_dat u,
ops_dat u0,
ops_dat p,
ops_dat r,
ops_dat Mi,
ops_dat w,
ops_dat z,
ops_dat Kx,
ops_dat Ky,
ops_dat cp,
ops_dat bfp,
double *ch_alphas,
double *ch_betas,
double rx, double ry, int step, int preconditioner_type)
{
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[] = {x_min,x_max,y_min,y_max};
ops_par_loop(tea_leaf_cheby_init_kernel, "tea_leaf_cheby_init_kernel", tea_grid, 2, rangexy,
ops_arg_dat(w, 1, S2D_00, "double", OPS_WRITE),
ops_arg_dat(r, 1, S2D_00, "double", OPS_WRITE),
ops_arg_dat(Kx, 1, S2D_00_P10, "double", OPS_READ),
ops_arg_dat(Ky, 1, S2D_00_0P1, "double", OPS_READ),
ops_arg_dat(u, 1, S2D_00_0M1_M10_P10_0P1, "double", OPS_READ),
ops_arg_dat(u0, 1, S2D_00, "double", OPS_READ),
ops_arg_gbl(&rx, 1, "double", OPS_READ),
ops_arg_gbl(&ry, 1, "double", OPS_READ));
if (preconditioner_type != TL_PREC_NONE) {
if (preconditioner_type == TL_PREC_JAC_BLOCK)
tea_block_solve(r, z, cp, bfp, Kx, Ky, rx, ry);
else if (preconditioner_type == TL_PREC_JAC_DIAG)
tea_diag_solve(r, z, Mi, Kx, Ky, rx, ry);
ops_par_loop(tea_leaf_axpby_kernel, "tea_leaf_axpby_kernel", tea_grid, 2, rangexy,
ops_arg_dat(p, 1, S2D_00, "double", OPS_RW),
ops_arg_dat(z, 1, S2D_00, "double", OPS_READ),
ops_arg_gbl(&ch_alphas[step], 1, "double", OPS_READ),
ops_arg_gbl(&ch_betas[step], 1, "double", OPS_READ));
} else {
ops_par_loop(tea_leaf_axpby_kernel, "tea_leaf_axpby_kernel", tea_grid, 2, rangexy,
ops_arg_dat(p, 1, S2D_00, "double", OPS_RW),
ops_arg_dat(r, 1, S2D_00, "double", OPS_READ),
ops_arg_gbl(&ch_alphas[step], 1, "double", OPS_READ),
ops_arg_gbl(&ch_betas[step], 1, "double", OPS_READ));
}
ops_par_loop(tea_leaf_xpy_kernel, "tea_leaf_xpy_kernel", tea_grid, 2, rangexy,
ops_arg_dat(u, 1, S2D_00, "double", OPS_INC),
ops_arg_dat(p, 1, S2D_00, "double", OPS_READ));
}
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 advec_cell(int sweep_number, int dir)
{
//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 z_min = field.z_min;
int z_max = field.z_max;
int rangexyz[] = {x_min-2,x_max+2,y_min-2,y_max+2,z_min-2,z_max+2}; // full range over grid
int rangexyz_inner[] = {x_min,x_max,y_min,y_max,z_min,z_max}; // inner range without border
int rangexyz_inner_plus2x[] = {x_min,x_max+2,y_min,y_max,z_min,z_max}; // inner range with +2 in x
int rangexyz_inner_plus2yz[] = {x_min,x_max,y_min,y_max+2,z_min,z_max+2}; // inner range with +2 in y and z
int rangexyz_inner_plus2z[] = {x_min,x_max,y_min,y_max,z_min,z_max+2}; // inner range with +2 in z
//printf("direction: %d sweep_number %d \n", dir, sweep_number);
if(dir == g_xdir) {
if(sweep_number == 1) {
ops_par_loop(advec_cell_kernel1_xdir, "advec_cell_kernel1_xdir", clover_grid, 3, rangexyz,
ops_arg_dat(work_array1, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array2, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(volume, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(vol_flux_x, 1, S3D_000_P100, "double", OPS_READ),
ops_arg_dat(vol_flux_y, 1, S3D_000_0P10, "double", OPS_READ),
ops_arg_dat(vol_flux_z, 1, S3D_000_00P1, "double", OPS_READ));
}
else if (sweep_number == 3) {
ops_par_loop(advec_cell_kernel2_xdir, "advec_cell_kernel2_xdir", clover_grid, 3, rangexyz,
ops_arg_dat(work_array1, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array2, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(volume, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(vol_flux_x, 1, S3D_000_P100, "double", OPS_READ));
}
ops_par_loop(advec_cell_kernel3_xdir, "advec_cell_kernel3_xdir", clover_grid, 3, rangexyz_inner_plus2x,
ops_arg_dat(vol_flux_x, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array1, 1, S3D_000_M100, "double", OPS_READ),
ops_arg_dat(xx, 1, S3D_000_P100_STRID3D_X, "int", OPS_READ),
ops_arg_dat(vertexdx, 1, S3D_000_P100_M100_STRID3D_X, "double", OPS_READ),
ops_arg_dat(density1, 1, S3D_000_P100_M100_M200, "double", OPS_READ),
ops_arg_dat(energy1, 1, S3D_000_P100_M100_M200, "double", OPS_READ),
ops_arg_dat(mass_flux_x, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, 1, S3D_000, "double", OPS_WRITE));
ops_par_loop(advec_cell_kernel4_xdir, "advec_cell_kernel4_xdir", clover_grid, 3, rangexyz_inner,
ops_arg_dat(density1, 1, S3D_000, "double", OPS_RW),
ops_arg_dat(energy1, 1, S3D_000, "double", OPS_RW),
ops_arg_dat(mass_flux_x, 1, S3D_000_P100, "double", OPS_READ),
ops_arg_dat(vol_flux_x, 1, S3D_000_P100, "double", OPS_READ),
ops_arg_dat(work_array1, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array2, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array3, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array4, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array5, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array6, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, 1, S3D_000_P100, "double", OPS_READ));
}
else if(dir == g_ydir) {
if(sweep_number == 2) {
if (advect_x) {
ops_par_loop(advec_cell_kernel1_ydir, "advec_cell_kernel1_ydir", clover_grid, 3, rangexyz,
ops_arg_dat(work_array1, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array2, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(volume, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(vol_flux_z, 1, S3D_000_00P1, "double", OPS_READ),
ops_arg_dat(vol_flux_y, 1, S3D_000_0P10, "double", OPS_READ));
}
else {
ops_par_loop(advec_cell_kernel2_ydir, "advec_cell_kernel2_ydir", clover_grid, 3, rangexyz,
ops_arg_dat(work_array1, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array2, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(volume, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(vol_flux_y, 1, S3D_000_0P10, "double", OPS_READ),
ops_arg_dat(vol_flux_x, 1, S3D_000_P100, "double", OPS_READ));
}
}
ops_par_loop(advec_cell_kernel3_ydir, "advec_cell_kernel3_ydir", clover_grid, 3, rangexyz_inner_plus2yz,
ops_arg_dat(vol_flux_y, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array1, 1, S3D_000_0M10, "double", OPS_READ),
ops_arg_dat(yy, 1, S3D_000_0P10_STRID3D_Y, "int", OPS_READ),
ops_arg_dat(vertexdy, 1, S3D_000_0P10_0M10_STRID3D_Y, "double", OPS_READ),
ops_arg_dat(density1, 1, S3D_000_0P10_0M10_0M20, "double", OPS_READ),
ops_arg_dat(energy1, 1, S3D_000_0P10_0M10_0M20, "double", OPS_READ),
ops_arg_dat(mass_flux_y, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, 1, S3D_000, "double", OPS_WRITE));
ops_par_loop(advec_cell_kernel4_ydir, "advec_cell_kernel4_ydir", clover_grid, 3, rangexyz_inner,
ops_arg_dat(density1, 1, S3D_000, "double", OPS_RW),
ops_arg_dat(energy1, 1, S3D_000, "double", OPS_RW),
ops_arg_dat(mass_flux_y, 1, S3D_000_0P10, "double", OPS_READ),
ops_arg_dat(vol_flux_y, 1, S3D_000_0P10, "double", OPS_READ),
ops_arg_dat(work_array1, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array2, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array3, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array4, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array5, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array6, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, 1, S3D_000_0P10, "double", OPS_READ));
}
else if(dir == g_zdir) {
if(sweep_number == 1) {
ops_par_loop(advec_cell_kernel1_zdir, "advec_cell_kernel1_zdir", clover_grid, 3, rangexyz,
ops_arg_dat(work_array1, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array2, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(volume, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(vol_flux_x, 1, S3D_000_P100, "double", OPS_READ),
ops_arg_dat(vol_flux_y, 1, S3D_000_0P10, "double", OPS_READ),
ops_arg_dat(vol_flux_z, 1, S3D_000_00P1, "double", OPS_READ));
}
else if (sweep_number == 3) {
ops_par_loop(advec_cell_kernel2_zdir, "advec_cell_kernel2_zdir", clover_grid, 3, rangexyz,
ops_arg_dat(work_array1, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array2, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(volume, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(vol_flux_z, 1, S3D_000_00P1, "double", OPS_READ));
}
ops_par_loop(advec_cell_kernel3_zdir, "advec_cell_kernel3_zdir", clover_grid, 3, rangexyz_inner_plus2z,
ops_arg_dat(vol_flux_z, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array1, 1, S3D_000_00M1, "double", OPS_READ),
ops_arg_dat(zz, 1, S3D_000_00P1_STRID3D_Z, "int", OPS_READ),
ops_arg_dat(vertexdz, 1, S3D_000_00P1_00M1_STRID3D_Z, "double", OPS_READ),
ops_arg_dat(density1, 1, S3D_000_00P1_00M1_00M2, "double", OPS_READ),
ops_arg_dat(energy1, 1, S3D_000_00P1_00M1_00M2, "double", OPS_READ),
ops_arg_dat(mass_flux_z, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, 1, S3D_000, "double", OPS_WRITE));
ops_par_loop(advec_cell_kernel4_zdir, "advec_cell_kernel4_zdir", clover_grid, 3, rangexyz_inner,
ops_arg_dat(density1, 1, S3D_000, "double", OPS_RW),
ops_arg_dat(energy1, 1, S3D_000, "double", OPS_RW),
ops_arg_dat(mass_flux_z, 1, S3D_000_00P1, "double", OPS_READ),
ops_arg_dat(vol_flux_z, 1, S3D_000_00P1, "double", OPS_READ),
ops_arg_dat(work_array1, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array2, 1, S3D_000, "double", OPS_READ),
ops_arg_dat(work_array3, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array4, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array5, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array6, 1, S3D_000, "double", OPS_WRITE),
ops_arg_dat(work_array7, 1, S3D_000_00P1, "double", OPS_READ));
}
}
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);
}
