1 // Copyright (c) 2017-2022, Lawrence Livermore National Security, LLC and other CEED contributors. 2 // All Rights Reserved. See the top-level LICENSE and NOTICE files for details. 3 // 4 // SPDX-License-Identifier: BSD-2-Clause 5 // 6 // This file is part of CEED: http://github.com/ceed 7 8 // libCEED + PETSc Example: Navier-Stokes 9 // 10 // This example demonstrates a simple usage of libCEED with PETSc to solve a Navier-Stokes problem. 11 // 12 // Build with: 13 // 14 // make [PETSC_DIR=</path/to/petsc>] [CEED_DIR=</path/to/libceed>] navierstokes 15 // 16 // Sample runs: 17 // 18 // ./navierstokes -ceed /cpu/self -problem density_current -degree 1 19 // ./navierstokes -ceed /gpu/cuda -problem advection -degree 1 20 // 21 //TESTARGS(name="blasius_aniso_diff_filter") -ceed {ceed_resource} -test_type diff_filter -options_file examples/fluids/tests-output/blasius_test.yaml -compare_final_state_atol 5e-10 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-blasius_diff_filter_aniso_vandriest.bin -diff_filter_monitor -ts_max_steps 0 -state_var primitive -diff_filter_friction_length 1e-5 -diff_filter_wall_damping_function van_driest -diff_filter_ksp_rtol 1e-8 -diff_filter_grid_based_width -diff_filter_width_scaling 1,0.7,1 22 //TESTARGS(name="blasius_iso_diff_filter") -ceed {ceed_resource} -test_type diff_filter -options_file examples/fluids/tests-output/blasius_test.yaml -compare_final_state_atol 2e-12 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-blasius_diff_filter_iso.bin -diff_filter_monitor -ts_max_steps 0 -diff_filter_width_scaling 4.2e-5,4.2e-5,4.2e-5 -diff_filter_ksp_atol 1e-14 -diff_filter_ksp_rtol 1e-16 23 //TESTARGS(name="blasius_SGS_DataDriven") -ceed {ceed_resource} -test_type solver -options_file examples/fluids/tests-output/blasius_stgtest.yaml -sgs_model_type data_driven -sgs_model_dd_leakyrelu_alpha 0.3 -sgs_model_dd_parameter_dir examples/fluids/dd_sgs_data -ts_dt 1e-9 -compare_final_state_atol 2e-12 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-blasius-sgs-data-driven.bin -state_var primitive 24 //TESTARGS(name="gaussianwave_idl") -ceed {ceed_resource} -test_type solver -options_file examples/fluids/gaussianwave.yaml -compare_final_state_atol 2e-11 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-gaussianwave-IDL.bin -dm_plex_box_faces 5,5,1 -ts_max_steps 5 -idl_decay_time 2e-3 -idl_length 0.25 -idl_start 0 25 //TESTARGS(name="turb_spanstats") -ceed {ceed_resource} -test_type turb_spanstats -options_file examples/fluids/tests-output/stats_test.yaml -compare_final_state_atol 1E-11 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-turb-spanstats-stats.bin 26 //TESTARGS(name="blasius") -ceed {ceed_resource} -test_type solver -options_file examples/fluids/tests-output/blasius_test.yaml -compare_final_state_atol 2E-11 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-blasius.bin 27 //TESTARGS(name="blasius_STG") -ceed {ceed_resource} -test_type solver -options_file examples/fluids/tests-output/blasius_stgtest.yaml -compare_final_state_atol 2E-11 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-blasius_STG.bin 28 //TESTARGS(name="blasius_STG_weakT") -ceed {ceed_resource} -test_type solver -options_file examples/fluids/tests-output/blasius_stgtest.yaml -compare_final_state_atol 1E-11 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-blasius_STG_weakT.bin -weakT 29 //TESTARGS(name="blasius_STG_strongBC") -ceed {ceed_resource} -test_type solver -options_file examples/fluids/tests-output/blasius_stgtest.yaml -compare_final_state_atol 1E-10 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-blasius_STG_strongBC.bin -stg_strong true 30 //TESTARGS(name="channel") -ceed {ceed_resource} -test_type solver -options_file examples/fluids/channel.yaml -compare_final_state_atol 2e-11 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-channel.bin -dm_plex_box_faces 5,5,1 -ts_max_steps 5 31 //TESTARGS(name="channel-primitive") -ceed {ceed_resource} -test_type solver -options_file examples/fluids/channel.yaml -compare_final_state_atol 2e-11 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-channel-prim.bin -dm_plex_box_faces 5,5,1 -ts_max_steps 5 -state_var primitive 32 //TESTARGS(name="dc_explicit") -ceed {ceed_resource} -test_type solver -degree 3 -q_extra 2 -dm_plex_box_faces 1,1,2 -dm_plex_box_lower 0,0,0 -dm_plex_box_upper 125,125,250 -dm_plex_dim 3 -bc_slip_x 5,6 -bc_slip_y 3,4 -bc_Slip_z 1,2 -units_kilogram 1e-9 -center 62.5,62.5,187.5 -rc 100. -thetaC -35. -mu 75 -ts_dt 1e-3 -units_meter 1e-2 -units_second 1e-2 -compare_final_state_atol 1E-11 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-dc-explicit.bin 33 //TESTARGS(name="dc_implicit_stab_none") -ceed {ceed_resource} -test_type solver -degree 3 -dm_plex_box_faces 1,1,2 -dm_plex_box_lower 0,0,0 -dm_plex_box_upper 125,125,250 -dm_plex_dim 3 -bc_slip_x 5,6 -bc_slip_y 3,4 -bc_Slip_z 1,2 -units_kilogram 1e-9 -center 62.5,62.5,187.5 -rc 100. -thetaC -35. -mu 75 -units_meter 1e-2 -units_second 1e-2 -ksp_atol 1e-4 -ksp_rtol 1e-3 -ksp_type bcgs -snes_atol 1e-3 -snes_lag_jacobian 100 -snes_lag_jacobian_persists -snes_mf_operator -ts_dt 1e-3 -implicit -ts_type alpha -compare_final_state_atol 5E-4 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-dc-implicit-stab-none.bin 34 //TESTARGS(name="adv_rotation_implicit_stab_supg") -ceed {ceed_resource} -test_type solver -problem advection -CtauS .3 -stab supg -degree 3 -dm_plex_box_faces 1,1,2 -dm_plex_box_lower 0,0,0 -dm_plex_box_upper 125,125,250 -dm_plex_dim 3 -bc_wall 1,2,3,4,5,6 -wall_comps 4 -units_kilogram 1e-9 -rc 100. -ksp_atol 1e-4 -ksp_rtol 1e-3 -ksp_type bcgs -snes_atol 1e-3 -snes_lag_jacobian 100 -snes_lag_jacobian_persists -snes_mf_operator -ts_dt 1e-3 -implicit -dm_mat_preallocate_skip 0 -ts_type alpha -compare_final_state_atol 5E-4 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-adv-rotation-implicit-stab-supg.bin 35 //TESTARGS(name="adv_translation_implicit_stab_su") -ceed {ceed_resource} -test_type solver -problem advection -CtauS .3 -stab su -degree 3 -dm_plex_box_faces 1,1,2 -dm_plex_box_lower 0,0,0 -dm_plex_box_upper 125,125,250 -dm_plex_dim 3 -units_kilogram 1e-9 -rc 100. -ksp_atol 1e-4 -ksp_rtol 1e-3 -ksp_type bcgs -snes_atol 1e-3 -snes_lag_jacobian 100 -snes_lag_jacobian_persists -snes_mf_operator -ts_dt 1e-3 -implicit -dm_mat_preallocate_skip 0 -ts_type alpha -wind_type translation -wind_translation .53,-1.33,-2.65 -bc_inflow 1,2,3,4,5,6 -compare_final_state_atol 5E-4 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-adv-translation-implicit-stab-su.bin 36 //TESTARGS(name="adv2d_rotation_explicit_strong") -ceed {ceed_resource} -test_type solver -problem advection2d -strong_form 1 -degree 3 -dm_plex_box_faces 2,2 -dm_plex_box_lower 0,0 -dm_plex_box_upper 125,125 -bc_wall 1,2,3,4 -wall_comps 4 -units_kilogram 1e-9 -rc 100. -ts_dt 1e-3 -compare_final_state_atol 1E-11 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-adv2d-rotation-explicit-strong.bin 37 //TESTARGS(name="adv2d_rotation_implicit_stab_supg") -ceed {ceed_resource} -test_type solver -problem advection2d -CtauS .3 -stab supg -degree 3 -dm_plex_box_faces 1,1,2 -dm_plex_box_lower 0,0 -dm_plex_box_upper 125,125 -bc_wall 1,2,3,4 -wall_comps 4 -units_kilogram 1e-9 -rc 100. -ksp_atol 1e-4 -ksp_rtol 1e-3 -ksp_type bcgs -snes_atol 1e-3 -snes_lag_jacobian 100 -snes_lag_jacobian_persists -snes_mf_operator -ts_dt 1e-3 -implicit -dm_mat_preallocate_skip 0 -ts_type alpha -compare_final_state_atol 5E-4 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-adv2d-rotation-implicit-stab-supg.bin 38 //TESTARGS(name="euler_implicit") -ceed {ceed_resource} -test_type solver -problem euler_vortex -degree 3 -dm_plex_box_faces 1,1,2 -dm_plex_box_lower 0,0,0 -dm_plex_box_upper 125,125,250 -dm_plex_dim 3 -units_meter 1e-4 -units_second 1e-4 -mean_velocity 1.4,-2.,0 -bc_inflow 4,6 -bc_outflow 3,5 -bc_slip_z 1,2 -vortex_strength 2 -ksp_atol 1e-4 -ksp_rtol 1e-3 -ksp_type bcgs -snes_atol 1e-3 -snes_lag_jacobian 100 -snes_lag_jacobian_persists -snes_mf_operator -ts_dt 1e-3 -implicit -dm_mat_preallocate_skip 0 -ts_type alpha -compare_final_state_atol 5E-4 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-euler-implicit.bin 39 //TESTARGS(name="euler_explicit") -ceed {ceed_resource} -test_type solver -problem euler_vortex -degree 3 -q_extra 2 -dm_plex_box_faces 2,2,1 -dm_plex_box_lower 0,0,0 -dm_plex_box_upper 125,125,250 -dm_plex_dim 3 -units_meter 1e-4 -units_second 1e-4 -mean_velocity 1.4,-2.,0 -bc_inflow 4,6 -bc_outflow 3,5 -bc_slip_z 1,2 -vortex_strength 2 -ts_dt 1e-7 -ts_rk_type 5bs -ts_rtol 1e-10 -ts_atol 1e-10 -compare_final_state_atol 1E-7 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-euler-explicit.bin 40 //TESTARGS(name="shocktube_explicit_su_yzb") -ceed {ceed_resource} -test_type solver -problem shocktube -degree 1 -q_extra 2 -dm_plex_box_faces 50,1,1 -units_meter 1e-2 units_second 1e-2 -dm_plex_box_lower 0,0,0 -dm_plex_box_upper 1000,20,20 -dm_plex_dim 3 -bc_slip_x 5,6 -bc_slip_y 3,4 -bc_Slip_z 1,2 -yzb -stab su -compare_final_state_atol 1E-11 -compare_final_state_filename examples/fluids/tests-output/fluids-navierstokes-shocktube-explicit-su-yzb.bin 41 42 /// @file 43 /// Navier-Stokes example using PETSc 44 45 #include "petscsf.h" 46 const char help[] = "Solve Navier-Stokes using PETSc and libCEED\n"; 47 48 #include <petscdevice.h> 49 #include "navierstokes.h" 50 51 #include <ceed.h> 52 #include <petscdmplex.h> 53 #include <petscts.h> 54 55 int main(int argc, char **argv) { 56 // --------------------------------------------------------------------------- 57 // Initialize PETSc 58 // --------------------------------------------------------------------------- 59 PetscCall(PetscInitialize(&argc, &argv, NULL, help)); 60 61 // --------------------------------------------------------------------------- 62 // Create structs 63 // --------------------------------------------------------------------------- 64 AppCtx app_ctx; 65 PetscCall(PetscCalloc1(1, &app_ctx)); 66 67 ProblemData *problem = NULL; 68 PetscCall(PetscCalloc1(1, &problem)); 69 70 User user; 71 PetscCall(PetscCalloc1(1, &user)); 72 73 CeedData ceed_data; 74 PetscCall(PetscCalloc1(1, &ceed_data)); 75 76 SimpleBC bc; 77 PetscCall(PetscCalloc1(1, &bc)); 78 79 Physics phys_ctx; 80 PetscCall(PetscCalloc1(1, &phys_ctx)); 81 82 Units units; 83 PetscCall(PetscCalloc1(1, &units)); 84 85 user->app_ctx = app_ctx; 86 user->units = units; 87 user->phys = phys_ctx; 88 problem->bc_from_ics = PETSC_TRUE; 89 90 PetscCall(RegisterLogEvents()); 91 92 // --------------------------------------------------------------------------- 93 // Process command line options 94 // --------------------------------------------------------------------------- 95 // -- Register problems to be available on the command line 96 PetscCall(RegisterProblems_NS(app_ctx)); 97 98 // -- Process general command line options 99 MPI_Comm comm = PETSC_COMM_WORLD; 100 user->comm = comm; 101 PetscCall(ProcessCommandLineOptions(comm, app_ctx, bc)); 102 103 // --------------------------------------------------------------------------- 104 // Initialize libCEED 105 // --------------------------------------------------------------------------- 106 // -- Initialize backend 107 Ceed ceed; 108 CeedInit(app_ctx->ceed_resource, &ceed); 109 user->ceed = ceed; 110 111 // -- Check preferred MemType 112 CeedMemType mem_type_backend; 113 CeedGetPreferredMemType(ceed, &mem_type_backend); 114 115 { 116 const char *resource; 117 CeedGetResource(ceed, &resource); 118 if (strstr(resource, "/gpu/sycl")) { 119 PetscDeviceContext dctx; 120 PetscCall(PetscDeviceContextGetCurrentContext(&dctx)); 121 void *stream_handle; 122 PetscCall(PetscDeviceContextGetStreamHandle(dctx, &stream_handle)); 123 CeedSetStream(ceed, stream_handle); 124 } 125 } 126 127 // --------------------------------------------------------------------------- 128 // Set up global mesh 129 // --------------------------------------------------------------------------- 130 // -- Create DM 131 DM dm; 132 VecType vec_type = NULL; 133 MatType mat_type = NULL; 134 switch (mem_type_backend) { 135 case CEED_MEM_HOST: 136 vec_type = VECSTANDARD; 137 break; 138 case CEED_MEM_DEVICE: { 139 const char *resolved; 140 CeedGetResource(ceed, &resolved); 141 if (strstr(resolved, "/gpu/cuda")) vec_type = VECCUDA; 142 else if (strstr(resolved, "/gpu/hip")) vec_type = VECKOKKOS; 143 else if (strstr(resolved, "/gpu/sycl")) vec_type = VECKOKKOS; 144 else vec_type = VECSTANDARD; 145 } 146 } 147 if (strstr(vec_type, VECCUDA)) mat_type = MATAIJCUSPARSE; 148 else if (strstr(vec_type, VECKOKKOS)) mat_type = MATAIJKOKKOS; 149 else mat_type = MATAIJ; 150 PetscCall(CreateDM(comm, problem, mat_type, vec_type, &dm)); 151 user->dm = dm; 152 PetscCall(DMSetApplicationContext(dm, user)); 153 154 // --------------------------------------------------------------------------- 155 // Choose the problem from the list of registered problems 156 // --------------------------------------------------------------------------- 157 { 158 PetscErrorCode (*p)(ProblemData *, DM, void *, SimpleBC); 159 PetscCall(PetscFunctionListFind(app_ctx->problems, app_ctx->problem_name, &p)); 160 PetscCheck(p, PETSC_COMM_SELF, 1, "Problem '%s' not found", app_ctx->problem_name); 161 PetscCall((*p)(problem, dm, &user, bc)); 162 } 163 164 // -- Set up DM 165 PetscCall(SetUpDM(dm, problem, app_ctx->degree, bc, phys_ctx)); 166 167 // -- Refine DM for high-order viz 168 if (app_ctx->viz_refine) PetscCall(VizRefineDM(dm, user, problem, bc, phys_ctx)); 169 170 // --------------------------------------------------------------------------- 171 // Create solution vectors 172 // --------------------------------------------------------------------------- 173 // -- Set up global state vector Q 174 Vec Q; 175 PetscCall(DMCreateGlobalVector(dm, &Q)); 176 PetscCall(VecZeroEntries(Q)); 177 178 // -- Set up local state vectors Q_loc, Q_dot_loc 179 PetscCall(DMCreateLocalVector(dm, &user->Q_loc)); 180 PetscCall(DMCreateLocalVector(dm, &user->Q_dot_loc)); 181 PetscCall(VecZeroEntries(user->Q_dot_loc)); 182 183 // --------------------------------------------------------------------------- 184 // Set up libCEED 185 // --------------------------------------------------------------------------- 186 // -- Set up libCEED objects 187 PetscCall(SetupLibceed(ceed, ceed_data, dm, user, app_ctx, problem, bc)); 188 189 // --------------------------------------------------------------------------- 190 // Set up ICs 191 // --------------------------------------------------------------------------- 192 // -- Fix multiplicity for ICs 193 PetscCall(ICs_FixMultiplicity(dm, ceed_data, user, user->Q_loc, Q, 0.0)); 194 195 // --------------------------------------------------------------------------- 196 // Set up lumped mass matrix 197 // --------------------------------------------------------------------------- 198 // -- Set up global mass vector 199 PetscCall(VecDuplicate(Q, &user->M_inv)); 200 201 // -- Compute lumped mass matrix 202 PetscCall(ComputeLumpedMassMatrix(ceed, dm, ceed_data, user->M_inv)); 203 204 // --------------------------------------------------------------------------- 205 // Record boundary values from initial condition 206 // --------------------------------------------------------------------------- 207 // -- This overrides DMPlexInsertBoundaryValues(). 208 // We use this for the main simulation DM because the reference DMPlexInsertBoundaryValues() is very slow on the GPU due to extra device-to-host 209 // communication. If we disable this, we should still get the same results due to the problem->bc function, but with potentially much slower 210 // execution. 211 if (problem->bc_from_ics) { 212 PetscCall(SetBCsFromICs_NS(dm, Q, user->Q_loc)); 213 } 214 215 // --------------------------------------------------------------------------- 216 // Create output directory 217 // --------------------------------------------------------------------------- 218 PetscMPIInt rank; 219 MPI_Comm_rank(comm, &rank); 220 if (!rank) { 221 PetscCall(PetscMkdir(app_ctx->output_dir)); 222 } 223 224 // --------------------------------------------------------------------------- 225 // Gather initial Q values in case of continuation of simulation 226 // --------------------------------------------------------------------------- 227 // -- Set up initial values from binary file 228 if (app_ctx->cont_steps) { 229 PetscCall(SetupICsFromBinary(comm, app_ctx, Q)); 230 } 231 232 // --------------------------------------------------------------------------- 233 // Print problem summary 234 // --------------------------------------------------------------------------- 235 if (app_ctx->test_type == TESTTYPE_NONE) { 236 // Header and rank 237 char host_name[PETSC_MAX_PATH_LEN]; 238 int comm_size; 239 PetscCall(PetscGetHostName(host_name, sizeof host_name)); 240 PetscCall(MPI_Comm_size(comm, &comm_size)); 241 PetscCall(PetscPrintf(comm, 242 "\n-- Navier-Stokes solver - libCEED + PETSc --\n" 243 " MPI:\n" 244 " Host Name : %s\n" 245 " Total ranks : %d\n", 246 host_name, comm_size)); 247 248 // Problem specific info 249 PetscCall(problem->print_info(problem, app_ctx)); 250 251 // libCEED 252 const char *used_resource; 253 CeedGetResource(ceed, &used_resource); 254 PetscCall(PetscPrintf(comm, 255 " libCEED:\n" 256 " libCEED Backend : %s\n" 257 " libCEED Backend MemType : %s\n", 258 used_resource, CeedMemTypes[mem_type_backend])); 259 // PETSc 260 char box_faces_str[PETSC_MAX_PATH_LEN] = "3,3,3"; 261 if (problem->dim == 2) box_faces_str[3] = '\0'; 262 PetscCall(PetscOptionsGetString(NULL, NULL, "-dm_plex_box_faces", box_faces_str, sizeof(box_faces_str), NULL)); 263 MatType mat_type; 264 VecType vec_type; 265 PetscCall(DMGetMatType(dm, &mat_type)); 266 PetscCall(DMGetVecType(dm, &vec_type)); 267 PetscCall(PetscPrintf(comm, 268 " PETSc:\n" 269 " Box Faces : %s\n" 270 " DM MatType : %s\n" 271 " DM VecType : %s\n" 272 " Time Stepping Scheme : %s\n", 273 box_faces_str, mat_type, vec_type, phys_ctx->implicit ? "implicit" : "explicit")); 274 if (app_ctx->cont_steps) { 275 PetscCall(PetscPrintf(comm, 276 " Continue:\n" 277 " Filename: : %s\n" 278 " Step: : %" PetscInt_FMT "\n" 279 " Time: : %g\n", 280 app_ctx->cont_file, app_ctx->cont_steps, app_ctx->cont_time)); 281 } 282 // Mesh 283 const PetscInt num_comp_q = 5; 284 PetscInt glob_dofs, owned_dofs, local_dofs; 285 const CeedInt num_P = app_ctx->degree + 1, num_Q = num_P + app_ctx->q_extra; 286 // -- Get global size 287 PetscCall(DMGetGlobalVectorInfo(user->dm, &owned_dofs, &glob_dofs, NULL)); 288 // -- Get local size 289 PetscCall(DMGetLocalVectorInfo(user->dm, &local_dofs, NULL, NULL)); 290 PetscCall(PetscPrintf(comm, 291 " Mesh:\n" 292 " Number of 1D Basis Nodes (P) : %" CeedInt_FMT "\n" 293 " Number of 1D Quadrature Points (Q) : %" CeedInt_FMT "\n" 294 " Global DoFs : %" PetscInt_FMT "\n" 295 " DoFs per node : %" PetscInt_FMT "\n" 296 " Global nodes : %" PetscInt_FMT "\n", 297 num_P, num_Q, glob_dofs, num_comp_q, glob_dofs / num_comp_q)); 298 // -- Get Partition Statistics 299 PetscCall(PetscPrintf(comm, " Partition: (min,max,median,ratio)\n")); 300 { 301 PetscInt *gather_buffer = NULL; 302 PetscInt part_owned_dofs[3], part_local_dofs[3], part_shared_dofs[3]; 303 PetscInt median_index = comm_size % 2 ? comm_size / 2 : comm_size / 2 - 1; 304 MPI_Datatype PetscInt_MPI; 305 PetscCall(PetscDataTypeToMPIDataType(PETSC_INT, &PetscInt_MPI)); 306 if (!rank) PetscCall(PetscMalloc1(comm_size, &gather_buffer)); 307 308 PetscCallMPI(MPI_Gather(&owned_dofs, 1, PetscInt_MPI, gather_buffer, 1, PetscInt_MPI, 0, comm)); 309 if (!rank) { 310 PetscCall(PetscSortInt(comm_size, gather_buffer)); 311 part_owned_dofs[0] = gather_buffer[0]; // min 312 part_owned_dofs[1] = gather_buffer[comm_size - 1]; // max 313 part_owned_dofs[2] = gather_buffer[median_index]; // median 314 PetscReal part_owned_dof_ratio = (PetscReal)part_owned_dofs[1] / (PetscReal)part_owned_dofs[2]; 315 PetscCall(PetscPrintf(comm, " Owned free nodes : %" PetscInt_FMT ", %" PetscInt_FMT ", %" PetscInt_FMT ", %f\n", 316 part_owned_dofs[0] / num_comp_q, part_owned_dofs[1] / num_comp_q, part_owned_dofs[2] / num_comp_q, 317 part_owned_dof_ratio)); 318 } 319 320 PetscCallMPI(MPI_Gather(&local_dofs, 1, PetscInt_MPI, gather_buffer, 1, PetscInt_MPI, 0, comm)); 321 if (!rank) { 322 PetscCall(PetscSortInt(comm_size, gather_buffer)); 323 part_local_dofs[0] = gather_buffer[0]; // min 324 part_local_dofs[1] = gather_buffer[comm_size - 1]; // max 325 part_local_dofs[2] = gather_buffer[median_index]; // median 326 PetscReal part_local_dof_ratio = (PetscReal)part_local_dofs[1] / (PetscReal)part_local_dofs[2]; 327 PetscCall(PetscPrintf(comm, " Local nodes : %" PetscInt_FMT ", %" PetscInt_FMT ", %" PetscInt_FMT ", %f\n", 328 part_local_dofs[0] / num_comp_q, part_local_dofs[1] / num_comp_q, part_local_dofs[2] / num_comp_q, 329 part_local_dof_ratio)); 330 } 331 332 PetscInt num_remote_roots_total = 0; 333 { 334 PetscSF sf; 335 PetscInt nrranks; 336 const PetscInt *roffset, *rmine, *rremote; 337 const PetscMPIInt *rranks; 338 PetscCall(DMGetSectionSF(user->dm, &sf)); 339 PetscCall(PetscSFGetRootRanks(sf, &nrranks, &rranks, &roffset, &rmine, &rremote)); 340 for (PetscInt i = 0; i < nrranks; i++) { 341 if (rranks[i] == rank) continue; // Ignore same-part global->local transfers 342 num_remote_roots_total += roffset[i + 1] - roffset[i]; 343 } 344 } 345 PetscCallMPI(MPI_Gather(&num_remote_roots_total, 1, PetscInt_MPI, gather_buffer, 1, PetscInt_MPI, 0, comm)); 346 if (!rank) { 347 PetscCall(PetscSortInt(comm_size, gather_buffer)); 348 part_shared_dofs[0] = gather_buffer[0]; // min 349 part_shared_dofs[1] = gather_buffer[comm_size - 1]; // max 350 part_shared_dofs[2] = gather_buffer[median_index]; // median 351 PetscReal part_shared_dof_ratio = (PetscReal)part_shared_dofs[1] / (PetscReal)part_shared_dofs[2]; 352 PetscCall(PetscPrintf(comm, " Shared nodes : %" PetscInt_FMT ", %" PetscInt_FMT ", %" PetscInt_FMT ", %f\n", 353 part_shared_dofs[0] / num_comp_q, part_shared_dofs[1] / num_comp_q, part_shared_dofs[2] / num_comp_q, 354 part_shared_dof_ratio)); 355 } 356 357 if (!rank) PetscCall(PetscFree(gather_buffer)); 358 } 359 360 PetscCall(PetscPrintf(comm, "(nodes == DoFs / %" PetscInt_FMT ")\n", num_comp_q)); 361 } 362 // -- Zero Q_loc 363 PetscCall(VecZeroEntries(user->Q_loc)); 364 365 // --------------------------------------------------------------------------- 366 // TS: Create, setup, and solve 367 // --------------------------------------------------------------------------- 368 TS ts; 369 PetscScalar final_time; 370 PetscCall(TSSolve_NS(dm, user, app_ctx, phys_ctx, &Q, &final_time, &ts)); 371 372 // --------------------------------------------------------------------------- 373 // Post-processing 374 // --------------------------------------------------------------------------- 375 PetscCall(PostProcess_NS(ts, ceed_data, dm, problem, user, Q, final_time)); 376 377 // --------------------------------------------------------------------------- 378 // Destroy libCEED objects 379 // --------------------------------------------------------------------------- 380 381 PetscCall(TurbulenceStatisticsDestroy(user, ceed_data)); 382 PetscCall(NodalProjectionDataDestroy(user->grad_velo_proj)); 383 PetscCall(SGS_DD_DataDestroy(user->sgs_dd_data)); 384 PetscCall(DifferentialFilterDataDestroy(user->diff_filter)); 385 386 // -- Vectors 387 CeedVectorDestroy(&ceed_data->x_coord); 388 CeedVectorDestroy(&ceed_data->q_data); 389 CeedVectorDestroy(&user->q_ceed); 390 CeedVectorDestroy(&user->q_dot_ceed); 391 CeedVectorDestroy(&user->g_ceed); 392 CeedVectorDestroy(&user->coo_values_amat); 393 CeedVectorDestroy(&user->coo_values_pmat); 394 395 // -- Bases 396 CeedBasisDestroy(&ceed_data->basis_q); 397 CeedBasisDestroy(&ceed_data->basis_x); 398 CeedBasisDestroy(&ceed_data->basis_xc); 399 CeedBasisDestroy(&ceed_data->basis_q_sur); 400 CeedBasisDestroy(&ceed_data->basis_x_sur); 401 402 // -- Restrictions 403 CeedElemRestrictionDestroy(&ceed_data->elem_restr_q); 404 CeedElemRestrictionDestroy(&ceed_data->elem_restr_x); 405 CeedElemRestrictionDestroy(&ceed_data->elem_restr_qd_i); 406 407 // Destroy QFunction contexts after using 408 // ToDo: Simplify tracked libCEED objects, smaller struct 409 { 410 CeedQFunctionContextDestroy(&problem->apply_inflow_jacobian.qfunction_context); 411 CeedQFunctionContextDestroy(&problem->apply_inflow_jacobian.qfunction_context); 412 CeedQFunctionContextDestroy(&problem->apply_outflow_jacobian.qfunction_context); 413 CeedQFunctionContextDestroy(&problem->apply_outflow_jacobian.qfunction_context); 414 CeedQFunctionContextDestroy(&problem->apply_freestream_jacobian.qfunction_context); 415 CeedQFunctionContextDestroy(&problem->apply_freestream_jacobian.qfunction_context); 416 CeedQFunctionContextDestroy(&problem->setup_sur.qfunction_context); 417 CeedQFunctionContextDestroy(&problem->setup_vol.qfunction_context); 418 CeedQFunctionContextDestroy(&problem->ics.qfunction_context); 419 CeedQFunctionContextDestroy(&problem->apply_vol_rhs.qfunction_context); 420 CeedQFunctionContextDestroy(&problem->apply_vol_ifunction.qfunction_context); 421 CeedQFunctionContextDestroy(&problem->apply_vol_ijacobian.qfunction_context); 422 } 423 424 // -- QFunctions 425 CeedQFunctionDestroy(&ceed_data->qf_setup_vol); 426 CeedQFunctionDestroy(&ceed_data->qf_ics); 427 CeedQFunctionDestroy(&ceed_data->qf_rhs_vol); 428 CeedQFunctionDestroy(&ceed_data->qf_ifunction_vol); 429 CeedQFunctionDestroy(&ceed_data->qf_setup_sur); 430 CeedQFunctionDestroy(&ceed_data->qf_apply_inflow); 431 CeedQFunctionDestroy(&ceed_data->qf_apply_inflow_jacobian); 432 CeedQFunctionDestroy(&ceed_data->qf_apply_freestream); 433 CeedQFunctionDestroy(&ceed_data->qf_apply_freestream_jacobian); 434 435 // -- Operators 436 CeedOperatorDestroy(&ceed_data->op_setup_vol); 437 PetscCall(OperatorApplyContextDestroy(ceed_data->op_ics_ctx)); 438 CeedOperatorDestroy(&user->op_rhs_vol); 439 CeedOperatorDestroy(&user->op_ifunction_vol); 440 PetscCall(OperatorApplyContextDestroy(user->op_rhs_ctx)); 441 CeedOperatorDestroy(&user->op_ifunction); 442 CeedOperatorDestroy(&user->op_ijacobian); 443 444 // -- Ceed 445 CeedDestroy(&ceed); 446 447 if (app_ctx->test_type != TESTTYPE_NONE) { 448 PetscInt num_options_left = 0; 449 PetscCall(PetscOptionsLeftGet(NULL, &num_options_left, NULL, NULL)); 450 PetscCheck(num_options_left == 0, PETSC_COMM_WORLD, -1, 451 "There are unused options. This is not allowed. See error message for the unused options (or use -options_left directly)"); 452 } 453 454 // --------------------------------------------------------------------------- 455 // Clean up PETSc 456 // --------------------------------------------------------------------------- 457 // -- Vectors 458 PetscCall(VecDestroy(&Q)); 459 PetscCall(VecDestroy(&user->M_inv)); 460 PetscCall(VecDestroy(&user->Q_loc)); 461 PetscCall(VecDestroy(&user->Q_dot_loc)); 462 463 // -- Matrices 464 PetscCall(MatDestroy(&user->interp_viz)); 465 466 // -- DM 467 PetscCall(DMDestroy(&dm)); 468 PetscCall(DMDestroy(&user->dm_viz)); 469 470 // -- TS 471 PetscCall(TSDestroy(&ts)); 472 473 // -- Function list 474 PetscCall(PetscFunctionListDestroy(&app_ctx->problems)); 475 476 PetscCall(PetscFree(app_ctx->amat_type)); 477 PetscCall(PetscFree(app_ctx->wall_forces.walls)); 478 PetscCall(PetscViewerDestroy(&app_ctx->wall_forces.viewer)); 479 480 // -- Structs 481 PetscCall(PetscFree(units)); 482 PetscCall(PetscFree(user)); 483 PetscCall(PetscFree(problem)); 484 PetscCall(PetscFree(bc)); 485 PetscCall(PetscFree(phys_ctx)); 486 PetscCall(PetscFree(app_ctx)); 487 PetscCall(PetscFree(ceed_data)); 488 489 return PetscFinalize(); 490 } 491