1 // Copyright (c) 2017-2024, 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: CEED BPs 3-6 with Multigrid 9 // 10 // This example demonstrates a simple usage of libCEED with PETSc to solve the CEED BP benchmark problems, see http://ceed.exascaleproject.org/bps. 11 // 12 // The code uses higher level communication protocols in DMPlex. 13 // 14 // Build with: 15 // 16 // make multigrid [PETSC_DIR=</path/to/petsc>] [CEED_DIR=</path/to/libceed>] 17 // 18 // Sample runs: 19 // 20 // multigrid -problem bp3 21 // multigrid -problem bp4 22 // multigrid -problem bp5 -ceed /cpu/self 23 // multigrid -problem bp6 -ceed /gpu/cuda 24 // 25 //TESTARGS(name="BP3, hex elements") -ceed {ceed_resource} -test -problem bp3 -degree 3 26 //TESTARGS(name="BP3, tet elements") -ceed {ceed_resource} -test -problem bp3 -degree 3 -simplex 27 28 /// @file 29 /// CEED BPs 1-6 multigrid example using PETSc 30 const char help[] = "Solve CEED BPs using p-multigrid with PETSc and DMPlex\n"; 31 32 #include <ceed.h> 33 #include <petsc.h> 34 #include <petscdmplex.h> 35 #include <petscksp.h> 36 #include <petscsys.h> 37 #include <stdbool.h> 38 #include <string.h> 39 40 #include "bps.h" 41 #include "include/bpsproblemdata.h" 42 #include "include/libceedsetup.h" 43 #include "include/matops.h" 44 #include "include/petscutils.h" 45 #include "include/petscversion.h" 46 #include "include/structs.h" 47 48 int main(int argc, char **argv) { 49 MPI_Comm comm; 50 char filename[PETSC_MAX_PATH_LEN], ceed_resource[PETSC_MAX_PATH_LEN] = "/cpu/self"; 51 double my_rt_start, my_rt, rt_min, rt_max; 52 PetscInt degree = 3, q_extra, *l_size, *xl_size, *g_size, dim = 3, fine_level, mesh_elem[3] = {3, 3, 3}, num_comp_u = 1, num_levels = degree, 53 *level_degrees; 54 PetscScalar eps = 1.0; 55 PetscBool test_mode, benchmark_mode, read_mesh, write_solution, simplex; 56 PetscLogStage solve_stage; 57 PetscLogEvent assemble_event; 58 DM *dm, dm_orig; 59 KSP ksp; 60 PC pc; 61 Mat *mat_O, *mat_pr, mat_coarse; 62 Vec *X, *X_loc, *mult, rhs, rhs_loc; 63 PetscMemType mem_type; 64 OperatorApplyContext *op_apply_ctx, op_error_ctx; 65 ProlongRestrContext *pr_restr_ctx; 66 Ceed ceed; 67 CeedData *ceed_data; 68 CeedVector rhs_ceed, target; 69 CeedQFunction qf_error; 70 CeedOperator op_error; 71 BPType bp_choice; 72 CoarsenType coarsen; 73 74 PetscCall(PetscInitialize(&argc, &argv, NULL, help)); 75 comm = PETSC_COMM_WORLD; 76 77 // Parse command line options 78 PetscOptionsBegin(comm, NULL, "CEED BPs in PETSc", NULL); 79 bp_choice = CEED_BP3; 80 PetscCall(PetscOptionsEnum("-problem", "CEED benchmark problem to solve", NULL, bp_types, (PetscEnum)bp_choice, (PetscEnum *)&bp_choice, NULL)); 81 num_comp_u = bp_options[bp_choice].num_comp_u; 82 test_mode = PETSC_FALSE; 83 PetscCall(PetscOptionsBool("-test", "Testing mode (do not print unless error is large)", NULL, test_mode, &test_mode, NULL)); 84 benchmark_mode = PETSC_FALSE; 85 PetscCall(PetscOptionsBool("-benchmark", "Benchmarking mode (prints benchmark statistics)", NULL, benchmark_mode, &benchmark_mode, NULL)); 86 write_solution = PETSC_FALSE; 87 PetscCall(PetscOptionsBool("-write_solution", "Write solution for visualization", NULL, write_solution, &write_solution, NULL)); 88 simplex = PETSC_FALSE; 89 PetscCall(PetscOptionsBool("-simplex", "Element topology (default:hex)", NULL, simplex, &simplex, NULL)); 90 if ((bp_choice == CEED_BP5 || bp_choice == CEED_BP6) && (simplex)) { 91 SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP, "BP5/6 is not supported with simplex"); 92 } 93 PetscCall(PetscOptionsScalar("-eps", "Epsilon parameter for Kershaw mesh transformation", NULL, eps, &eps, NULL)); 94 if (eps > 1 || eps <= 0) SETERRQ(PETSC_COMM_WORLD, PETSC_ERR_ARG_OUTOFRANGE, "-eps %g must be (0,1]", (double)PetscRealPart(eps)); 95 degree = test_mode ? 3 : 2; 96 PetscCall(PetscOptionsInt("-degree", "Polynomial degree of tensor product basis", NULL, degree, °ree, NULL)); 97 if (degree < 1) SETERRQ(PETSC_COMM_WORLD, PETSC_ERR_ARG_OUTOFRANGE, "-degree %" PetscInt_FMT " must be at least 1", degree); 98 q_extra = bp_options[bp_choice].q_extra; 99 PetscCall(PetscOptionsInt("-q_extra", "Number of extra quadrature points", NULL, q_extra, &q_extra, NULL)); 100 PetscCall(PetscOptionsString("-ceed", "CEED resource specifier", NULL, ceed_resource, ceed_resource, sizeof(ceed_resource), NULL)); 101 coarsen = COARSEN_UNIFORM; 102 PetscCall(PetscOptionsEnum("-coarsen", "Coarsening strategy to use", NULL, coarsen_types, (PetscEnum)coarsen, (PetscEnum *)&coarsen, NULL)); 103 read_mesh = PETSC_FALSE; 104 PetscCall(PetscOptionsString("-mesh", "Read mesh from file", NULL, filename, filename, sizeof(filename), &read_mesh)); 105 if (!read_mesh) { 106 PetscInt tmp = dim; 107 PetscCall(PetscOptionsIntArray("-cells", "Number of cells per dimension", NULL, mesh_elem, &tmp, NULL)); 108 } 109 PetscOptionsEnd(); 110 111 // Set up libCEED 112 CeedInit(ceed_resource, &ceed); 113 CeedMemType mem_type_backend; 114 CeedGetPreferredMemType(ceed, &mem_type_backend); 115 116 // Setup DM 117 if (read_mesh) { 118 PetscCall(DMPlexCreateFromFile(PETSC_COMM_WORLD, filename, NULL, PETSC_TRUE, &dm_orig)); 119 } else { 120 PetscCall(DMPlexCreateBoxMesh(PETSC_COMM_WORLD, dim, simplex, mesh_elem, NULL, NULL, NULL, PETSC_TRUE, 0, PETSC_FALSE, &dm_orig)); 121 } 122 123 // Set mesh vec_type 124 VecType vec_type = VECSTANDARD; 125 126 switch (mem_type_backend) { 127 case CEED_MEM_HOST: 128 vec_type = VECSTANDARD; 129 break; 130 case CEED_MEM_DEVICE: { 131 const char *resolved; 132 133 CeedGetResource(ceed, &resolved); 134 if (strstr(resolved, "/gpu/cuda")) vec_type = VECCUDA; 135 else if (strstr(resolved, "/gpu/hip/occa")) vec_type = VECSTANDARD; // https://github.com/CEED/libCEED/issues/678 136 else if (strstr(resolved, "/gpu/hip")) vec_type = VECHIP; 137 else vec_type = VECSTANDARD; 138 } 139 } 140 PetscCall(DMSetVecType(dm_orig, vec_type)); 141 PetscCall(DMSetFromOptions(dm_orig)); 142 PetscCall(DMViewFromOptions(dm_orig, NULL, "-dm_view")); 143 144 // Apply Kershaw mesh transformation 145 PetscCall(Kershaw(dm_orig, eps)); 146 147 // Allocate arrays for PETSc objects for each level 148 switch (coarsen) { 149 case COARSEN_UNIFORM: 150 num_levels = degree; 151 break; 152 case COARSEN_LOGARITHMIC: 153 num_levels = ceil(log(degree) / log(2)) + 1; 154 break; 155 } 156 PetscCall(PetscMalloc1(num_levels, &level_degrees)); 157 fine_level = num_levels - 1; 158 159 switch (coarsen) { 160 case COARSEN_UNIFORM: 161 for (PetscInt i = 0; i < num_levels; i++) level_degrees[i] = i + 1; 162 break; 163 case COARSEN_LOGARITHMIC: 164 for (PetscInt i = 0; i < num_levels - 1; i++) level_degrees[i] = pow(2, i); 165 level_degrees[fine_level] = degree; 166 break; 167 } 168 PetscCall(PetscMalloc1(num_levels, &dm)); 169 PetscCall(PetscMalloc1(num_levels, &X)); 170 PetscCall(PetscMalloc1(num_levels, &X_loc)); 171 PetscCall(PetscMalloc1(num_levels, &mult)); 172 PetscCall(PetscMalloc1(num_levels, &op_apply_ctx)); 173 PetscCall(PetscMalloc1(num_levels, &pr_restr_ctx)); 174 PetscCall(PetscMalloc1(num_levels, &mat_O)); 175 PetscCall(PetscMalloc1(num_levels, &mat_pr)); 176 PetscCall(PetscMalloc1(num_levels, &l_size)); 177 PetscCall(PetscMalloc1(num_levels, &xl_size)); 178 PetscCall(PetscMalloc1(num_levels, &g_size)); 179 180 PetscInt c_start, c_end; 181 PetscCall(DMPlexGetHeightStratum(dm_orig, 0, &c_start, &c_end)); 182 DMPolytopeType cell_type; 183 PetscCall(DMPlexGetCellType(dm_orig, c_start, &cell_type)); 184 CeedElemTopology elem_topo = ElemTopologyP2C(cell_type); 185 186 // Setup DM and Operator Mat Shells for each level 187 for (PetscInt i = 0; i < num_levels; i++) { 188 // Create DM 189 PetscCall(DMClone(dm_orig, &dm[i])); 190 PetscCall(DMGetVecType(dm_orig, &vec_type)); 191 PetscCall(DMSetVecType(dm[i], vec_type)); 192 PetscInt dim; 193 PetscCall(DMGetDimension(dm[i], &dim)); 194 PetscCall(SetupDMByDegree(dm[i], level_degrees[fine_level], q_extra, num_comp_u, dim, bp_options[bp_choice].enforce_bc)); 195 196 // Create vectors 197 PetscCall(DMCreateGlobalVector(dm[i], &X[i])); 198 PetscCall(VecGetLocalSize(X[i], &l_size[i])); 199 PetscCall(VecGetSize(X[i], &g_size[i])); 200 PetscCall(DMCreateLocalVector(dm[i], &X_loc[i])); 201 PetscCall(VecGetSize(X_loc[i], &xl_size[i])); 202 203 // Operator 204 PetscCall(PetscMalloc1(1, &op_apply_ctx[i])); 205 PetscCall(MatCreateShell(comm, l_size[i], l_size[i], g_size[i], g_size[i], op_apply_ctx[i], &mat_O[i])); 206 PetscCall(MatShellSetOperation(mat_O[i], MATOP_MULT, (void (*)(void))MatMult_Ceed)); 207 PetscCall(MatShellSetOperation(mat_O[i], MATOP_GET_DIAGONAL, (void (*)(void))MatGetDiag)); 208 PetscCall(MatShellSetVecType(mat_O[i], vec_type)); 209 210 // Level transfers 211 if (i > 0) { 212 // Interp 213 PetscCall(PetscMalloc1(1, &pr_restr_ctx[i])); 214 PetscCall(MatCreateShell(comm, l_size[i], l_size[i - 1], g_size[i], g_size[i - 1], pr_restr_ctx[i], &mat_pr[i])); 215 PetscCall(MatShellSetOperation(mat_pr[i], MATOP_MULT, (void (*)(void))MatMult_Prolong)); 216 PetscCall(MatShellSetOperation(mat_pr[i], MATOP_MULT_TRANSPOSE, (void (*)(void))MatMult_Restrict)); 217 PetscCall(MatShellSetVecType(mat_pr[i], vec_type)); 218 } 219 } 220 PetscCall(VecDuplicate(X[fine_level], &rhs)); 221 222 // Print global grid information 223 if (!test_mode) { 224 PetscInt P = degree + 1, Q = P + q_extra; 225 226 const char *used_resource; 227 CeedGetResource(ceed, &used_resource); 228 229 PetscCall(VecGetType(X[0], &vec_type)); 230 231 PetscCall(PetscPrintf(comm, 232 "\n-- CEED Benchmark Problem %" CeedInt_FMT " -- libCEED + PETSc + PCMG --\n" 233 " PETSc:\n" 234 " PETSc Vec Type : %s\n" 235 " libCEED:\n" 236 " libCEED Backend : %s\n" 237 " libCEED Backend MemType : %s\n" 238 " Mesh:\n" 239 " Solution Order (P) : %" PetscInt_FMT "\n" 240 " Quadrature Order (Q) : %" PetscInt_FMT "\n" 241 " Additional quadrature points (q_extra) : %" PetscInt_FMT "\n" 242 " Global Nodes : %" PetscInt_FMT "\n" 243 " Owned Nodes : %" PetscInt_FMT "\n" 244 " DoF per node : %" PetscInt_FMT "\n" 245 " Element topology : %s\n" 246 " Multigrid:\n" 247 " Number of Levels : %" PetscInt_FMT "\n", 248 bp_choice + 1, vec_type, used_resource, CeedMemTypes[mem_type_backend], P, Q, q_extra, g_size[fine_level] / num_comp_u, 249 l_size[fine_level] / num_comp_u, num_comp_u, CeedElemTopologies[elem_topo], num_levels)); 250 } 251 252 // Create RHS vector 253 PetscCall(VecDuplicate(X_loc[fine_level], &rhs_loc)); 254 PetscCall(VecZeroEntries(rhs_loc)); 255 CeedVectorCreate(ceed, xl_size[fine_level], &rhs_ceed); 256 PetscCall(VecP2C(rhs_loc, &mem_type, rhs_ceed)); 257 258 // Set up libCEED operators on each level 259 PetscCall(PetscMalloc1(num_levels, &ceed_data)); 260 for (PetscInt i = 0; i < num_levels; i++) { 261 // Print level information 262 if (!test_mode && (i == 0 || i == fine_level)) { 263 PetscCall(PetscPrintf(comm, 264 " Level %" PetscInt_FMT " (%s):\n" 265 " Solution Order (P) : %" PetscInt_FMT "\n" 266 " Global Nodes : %" PetscInt_FMT "\n" 267 " Owned Nodes : %" PetscInt_FMT "\n", 268 i, (i ? "fine" : "coarse"), level_degrees[i] + 1, g_size[i] / num_comp_u, l_size[i] / num_comp_u)); 269 } 270 PetscCall(PetscMalloc1(1, &ceed_data[i])); 271 PetscCall(SetupLibceedByDegree(dm[i], ceed, level_degrees[i], dim, q_extra, dim, num_comp_u, g_size[i], xl_size[i], bp_options[bp_choice], 272 ceed_data[i], i == fine_level, i == fine_level, rhs_ceed, &target)); 273 } 274 275 // Gather RHS 276 PetscCall(VecC2P(rhs_ceed, mem_type, rhs_loc)); 277 PetscCall(VecZeroEntries(rhs)); 278 PetscCall(DMLocalToGlobal(dm[fine_level], rhs_loc, ADD_VALUES, rhs)); 279 CeedVectorDestroy(&rhs_ceed); 280 281 // Create the error QFunction 282 CeedQFunctionCreateInterior(ceed, 1, bp_options[bp_choice].error, bp_options[bp_choice].error_loc, &qf_error); 283 CeedQFunctionAddInput(qf_error, "u", num_comp_u, CEED_EVAL_INTERP); 284 CeedQFunctionAddInput(qf_error, "true_soln", num_comp_u, CEED_EVAL_NONE); 285 CeedQFunctionAddInput(qf_error, "qdata", ceed_data[fine_level]->q_data_size, CEED_EVAL_NONE); 286 CeedQFunctionAddOutput(qf_error, "error", num_comp_u, CEED_EVAL_INTERP); 287 288 // Create the error operator 289 CeedOperatorCreate(ceed, qf_error, CEED_QFUNCTION_NONE, CEED_QFUNCTION_NONE, &op_error); 290 CeedOperatorSetField(op_error, "u", ceed_data[fine_level]->elem_restr_u, ceed_data[fine_level]->basis_u, CEED_VECTOR_ACTIVE); 291 CeedOperatorSetField(op_error, "true_soln", ceed_data[fine_level]->elem_restr_u_i, CEED_BASIS_NONE, target); 292 CeedOperatorSetField(op_error, "qdata", ceed_data[fine_level]->elem_restr_qd_i, CEED_BASIS_NONE, ceed_data[fine_level]->q_data); 293 CeedOperatorSetField(op_error, "error", ceed_data[fine_level]->elem_restr_u, ceed_data[fine_level]->basis_u, CEED_VECTOR_ACTIVE); 294 295 // Calculate multiplicity 296 for (PetscInt i = 0; i < num_levels; i++) { 297 PetscMemType mem_type; 298 299 // CEED vector 300 PetscCall(VecZeroEntries(X_loc[i])); 301 PetscCall(VecP2C(X_loc[i], &mem_type, ceed_data[i]->x_ceed)); 302 303 // Multiplicity 304 CeedElemRestrictionGetMultiplicity(ceed_data[i]->elem_restr_u, ceed_data[i]->x_ceed); 305 CeedVectorSyncArray(ceed_data[i]->x_ceed, CEED_MEM_HOST); 306 307 // Restore vector 308 PetscCall(VecC2P(ceed_data[i]->x_ceed, mem_type, X_loc[i])); 309 310 // Creat mult vector 311 PetscCall(VecDuplicate(X_loc[i], &mult[i])); 312 313 // Local-to-global 314 PetscCall(VecZeroEntries(X[i])); 315 PetscCall(DMLocalToGlobal(dm[i], X_loc[i], ADD_VALUES, X[i])); 316 PetscCall(VecZeroEntries(X_loc[i])); 317 318 // Global-to-local 319 PetscCall(DMGlobalToLocal(dm[i], X[i], INSERT_VALUES, mult[i])); 320 PetscCall(VecZeroEntries(X[i])); 321 322 // Multiplicity scaling 323 PetscCall(VecReciprocal(mult[i])); 324 } 325 326 // Set up Mat 327 for (PetscInt i = fine_level; i >= 0; i--) { 328 // Set up apply operator context 329 PetscCall(SetupApplyOperatorCtx(comm, dm[i], ceed, ceed_data[i], X_loc[i], op_apply_ctx[i])); 330 331 if (i > 0) { 332 // Prolongation/Restriction Operator 333 PetscCall(CeedLevelTransferSetup(dm[i - 1], ceed, i, num_comp_u, ceed_data, bp_options[bp_choice], mult[i])); 334 pr_restr_ctx[i]->comm = comm; 335 pr_restr_ctx[i]->dmf = dm[i]; 336 pr_restr_ctx[i]->dmc = dm[i - 1]; 337 pr_restr_ctx[i]->loc_vec_c = X_loc[i - 1]; 338 pr_restr_ctx[i]->loc_vec_f = op_apply_ctx[i]->Y_loc; 339 pr_restr_ctx[i]->mult_vec = mult[i]; 340 pr_restr_ctx[i]->ceed_vec_c = ceed_data[i - 1]->x_ceed; 341 pr_restr_ctx[i]->ceed_vec_f = ceed_data[i]->y_ceed; 342 pr_restr_ctx[i]->op_prolong = ceed_data[i]->op_prolong; 343 pr_restr_ctx[i]->op_restrict = ceed_data[i]->op_restrict; 344 pr_restr_ctx[i]->ceed = ceed; 345 } 346 } 347 348 // Assemble coarse grid Jacobian for AMG (or other sparse matrix) solve 349 PetscCall(DMCreateMatrix(dm[0], &mat_coarse)); 350 351 PetscCall(PetscLogEventRegister("AssembleMatrix", MAT_CLASSID, &assemble_event)); 352 { 353 // Assemble matrix analytically 354 PetscCount num_entries; 355 CeedInt *rows_ceed, *cols_ceed; 356 PetscInt *rows_petsc, *cols_petsc; 357 ISLocalToGlobalMapping ltog_row, ltog_col; 358 CeedVector coo_values; 359 360 CeedOperatorLinearAssembleSymbolic(op_apply_ctx[0]->op, &num_entries, &rows_ceed, &cols_ceed); 361 PetscCall(IntArrayCeedToPetsc(num_entries, &rows_ceed, &rows_petsc)); 362 PetscCall(IntArrayCeedToPetsc(num_entries, &cols_ceed, &cols_petsc)); 363 PetscCall(MatGetLocalToGlobalMapping(mat_coarse, <og_row, <og_col)); 364 PetscCall(ISLocalToGlobalMappingApply(ltog_row, num_entries, rows_petsc, rows_petsc)); 365 PetscCall(ISLocalToGlobalMappingApply(ltog_col, num_entries, cols_petsc, cols_petsc)); 366 PetscCall(MatSetPreallocationCOO(mat_coarse, num_entries, rows_petsc, cols_petsc)); 367 free(rows_petsc); 368 free(cols_petsc); 369 CeedVectorCreate(ceed, num_entries, &coo_values); 370 PetscCall(PetscLogEventBegin(assemble_event, mat_coarse, 0, 0, 0)); 371 CeedOperatorLinearAssemble(op_apply_ctx[0]->op, coo_values); 372 const CeedScalar *values; 373 CeedVectorGetArrayRead(coo_values, CEED_MEM_HOST, &values); 374 PetscCall(MatSetValuesCOO(mat_coarse, values, ADD_VALUES)); 375 CeedVectorRestoreArrayRead(coo_values, &values); 376 PetscCall(PetscLogEventEnd(assemble_event, mat_coarse, 0, 0, 0)); 377 CeedVectorDestroy(&coo_values); 378 } 379 380 // Set up KSP 381 PetscCall(KSPCreate(comm, &ksp)); 382 { 383 PetscCall(KSPSetType(ksp, KSPCG)); 384 PetscCall(KSPSetNormType(ksp, KSP_NORM_NATURAL)); 385 PetscCall(KSPSetTolerances(ksp, 1e-10, PETSC_DEFAULT, PETSC_DEFAULT, PETSC_DEFAULT)); 386 } 387 PetscCall(KSPSetFromOptions(ksp)); 388 PetscCall(KSPSetOperators(ksp, mat_O[fine_level], mat_O[fine_level])); 389 390 // Set up PCMG 391 PetscCall(KSPGetPC(ksp, &pc)); 392 PCMGCycleType pcmg_cycle_type = PC_MG_CYCLE_V; 393 { 394 PetscCall(PCSetType(pc, PCMG)); 395 396 // PCMG levels 397 PetscCall(PCMGSetLevels(pc, num_levels, NULL)); 398 for (PetscInt i = 0; i < num_levels; i++) { 399 // Smoother 400 KSP smoother; 401 PC smoother_pc; 402 PetscCall(PCMGGetSmoother(pc, i, &smoother)); 403 PetscCall(KSPSetType(smoother, KSPCHEBYSHEV)); 404 PetscCall(KSPChebyshevEstEigSet(smoother, 0, 0.1, 0, 1.1)); 405 PetscCall(KSPChebyshevEstEigSetUseNoisy(smoother, PETSC_TRUE)); 406 PetscCall(KSPSetOperators(smoother, mat_O[i], mat_O[i])); 407 PetscCall(KSPGetPC(smoother, &smoother_pc)); 408 PetscCall(PCSetType(smoother_pc, PCJACOBI)); 409 PetscCall(PCJacobiSetType(smoother_pc, PC_JACOBI_DIAGONAL)); 410 411 // Work vector 412 if (i < num_levels - 1) { 413 PetscCall(PCMGSetX(pc, i, X[i])); 414 } 415 416 // Level transfers 417 if (i > 0) { 418 // Interpolation 419 PetscCall(PCMGSetInterpolation(pc, i, mat_pr[i])); 420 } 421 422 // Coarse solve 423 KSP coarse; 424 PC coarse_pc; 425 PetscCall(PCMGGetCoarseSolve(pc, &coarse)); 426 PetscCall(KSPSetType(coarse, KSPPREONLY)); 427 PetscCall(KSPSetOperators(coarse, mat_coarse, mat_coarse)); 428 429 PetscCall(KSPGetPC(coarse, &coarse_pc)); 430 PetscCall(PCSetType(coarse_pc, PCGAMG)); 431 432 PetscCall(KSPSetOptionsPrefix(coarse, "coarse_")); 433 PetscCall(PCSetOptionsPrefix(coarse_pc, "coarse_")); 434 PetscCall(KSPSetFromOptions(coarse)); 435 PetscCall(PCSetFromOptions(coarse_pc)); 436 } 437 438 // PCMG options 439 PetscCall(PCMGSetType(pc, PC_MG_MULTIPLICATIVE)); 440 PetscCall(PCMGSetNumberSmooth(pc, 3)); 441 PetscCall(PCMGSetCycleType(pc, pcmg_cycle_type)); 442 } 443 444 // First run, if benchmarking 445 if (benchmark_mode) { 446 PetscCall(KSPSetTolerances(ksp, 1e-10, PETSC_DEFAULT, PETSC_DEFAULT, 1)); 447 PetscCall(VecZeroEntries(X[fine_level])); 448 my_rt_start = MPI_Wtime(); 449 PetscCall(KSPSolve(ksp, rhs, X[fine_level])); 450 my_rt = MPI_Wtime() - my_rt_start; 451 PetscCall(MPI_Allreduce(MPI_IN_PLACE, &my_rt, 1, MPI_DOUBLE, MPI_MIN, comm)); 452 // Set maxits based on first iteration timing 453 if (my_rt > 0.02) { 454 PetscCall(KSPSetTolerances(ksp, 1e-10, PETSC_DEFAULT, PETSC_DEFAULT, 5)); 455 } else { 456 PetscCall(KSPSetTolerances(ksp, 1e-10, PETSC_DEFAULT, PETSC_DEFAULT, 20)); 457 } 458 } 459 460 // Timed solve 461 PetscCall(VecZeroEntries(X[fine_level])); 462 PetscCall(PetscBarrier((PetscObject)ksp)); 463 464 // -- Performance logging 465 PetscCall(PetscLogStageRegister("Solve Stage", &solve_stage)); 466 PetscCall(PetscLogStagePush(solve_stage)); 467 468 // -- Solve 469 my_rt_start = MPI_Wtime(); 470 PetscCall(KSPSolve(ksp, rhs, X[fine_level])); 471 my_rt = MPI_Wtime() - my_rt_start; 472 473 // -- Performance logging 474 PetscCall(PetscLogStagePop()); 475 476 // Output results 477 { 478 KSPType ksp_type; 479 PCMGType pcmg_type; 480 KSPConvergedReason reason; 481 PetscReal rnorm; 482 PetscInt its; 483 PetscCall(KSPGetType(ksp, &ksp_type)); 484 PetscCall(KSPGetConvergedReason(ksp, &reason)); 485 PetscCall(KSPGetIterationNumber(ksp, &its)); 486 PetscCall(KSPGetResidualNorm(ksp, &rnorm)); 487 PetscCall(PCMGGetType(pc, &pcmg_type)); 488 if (!test_mode || reason < 0 || rnorm > 1e-8) { 489 PetscCall(PetscPrintf(comm, 490 " KSP:\n" 491 " KSP Type : %s\n" 492 " KSP Convergence : %s\n" 493 " Total KSP Iterations : %" PetscInt_FMT "\n" 494 " Final rnorm : %e\n", 495 ksp_type, KSPConvergedReasons[reason], its, (double)rnorm)); 496 PetscCall(PetscPrintf(comm, 497 " PCMG:\n" 498 " PCMG Type : %s\n" 499 " PCMG Cycle Type : %s\n", 500 PCMGTypes[pcmg_type], PCMGCycleTypes[pcmg_cycle_type])); 501 } 502 if (!test_mode) { 503 PetscCall(PetscPrintf(comm, " Performance:\n")); 504 } 505 { 506 // Set up error operator context 507 PetscCall(PetscMalloc1(1, &op_error_ctx)); 508 PetscCall(SetupErrorOperatorCtx(comm, dm[fine_level], ceed, ceed_data[fine_level], X_loc[fine_level], op_error, op_error_ctx)); 509 PetscScalar l2_error; 510 PetscCall(ComputeL2Error(X[fine_level], &l2_error, op_error_ctx)); 511 PetscReal tol = 5e-2; 512 if (!test_mode || l2_error > tol) { 513 PetscCall(MPI_Allreduce(&my_rt, &rt_min, 1, MPI_DOUBLE, MPI_MIN, comm)); 514 PetscCall(MPI_Allreduce(&my_rt, &rt_max, 1, MPI_DOUBLE, MPI_MAX, comm)); 515 PetscCall(PetscPrintf(comm, 516 " L2 Error : %e\n" 517 " CG Solve Time : %g (%g) sec\n", 518 (double)l2_error, rt_max, rt_min)); 519 } 520 } 521 if (benchmark_mode && (!test_mode)) { 522 PetscCall(PetscPrintf(comm, " DoFs/Sec in CG : %g (%g) million\n", 1e-6 * g_size[fine_level] * its / rt_max, 523 1e-6 * g_size[fine_level] * its / rt_min)); 524 } 525 } 526 527 if (write_solution) { 528 PetscViewer vtk_viewer_soln; 529 530 PetscCall(PetscViewerCreate(comm, &vtk_viewer_soln)); 531 PetscCall(PetscViewerSetType(vtk_viewer_soln, PETSCVIEWERVTK)); 532 PetscCall(PetscViewerFileSetName(vtk_viewer_soln, "solution.vtu")); 533 PetscCall(VecView(X[fine_level], vtk_viewer_soln)); 534 PetscCall(PetscViewerDestroy(&vtk_viewer_soln)); 535 } 536 537 // Cleanup 538 for (PetscInt i = 0; i < num_levels; i++) { 539 PetscCall(VecDestroy(&X[i])); 540 PetscCall(VecDestroy(&X_loc[i])); 541 PetscCall(VecDestroy(&mult[i])); 542 PetscCall(VecDestroy(&op_apply_ctx[i]->Y_loc)); 543 PetscCall(MatDestroy(&mat_O[i])); 544 PetscCall(PetscFree(op_apply_ctx[i])); 545 if (i > 0) { 546 PetscCall(MatDestroy(&mat_pr[i])); 547 PetscCall(PetscFree(pr_restr_ctx[i])); 548 } 549 PetscCall(CeedDataDestroy(i, ceed_data[i])); 550 PetscCall(DMDestroy(&dm[i])); 551 } 552 PetscCall(PetscFree(level_degrees)); 553 PetscCall(PetscFree(dm)); 554 PetscCall(PetscFree(X)); 555 PetscCall(PetscFree(X_loc)); 556 PetscCall(VecDestroy(&op_error_ctx->Y_loc)); 557 PetscCall(PetscFree(mult)); 558 PetscCall(PetscFree(mat_O)); 559 PetscCall(PetscFree(mat_pr)); 560 PetscCall(PetscFree(ceed_data)); 561 PetscCall(PetscFree(op_apply_ctx)); 562 PetscCall(PetscFree(op_error_ctx)); 563 PetscCall(PetscFree(pr_restr_ctx)); 564 PetscCall(PetscFree(l_size)); 565 PetscCall(PetscFree(xl_size)); 566 PetscCall(PetscFree(g_size)); 567 PetscCall(VecDestroy(&rhs)); 568 PetscCall(VecDestroy(&rhs_loc)); 569 PetscCall(MatDestroy(&mat_coarse)); 570 PetscCall(KSPDestroy(&ksp)); 571 PetscCall(DMDestroy(&dm_orig)); 572 CeedVectorDestroy(&target); 573 CeedQFunctionDestroy(&qf_error); 574 CeedOperatorDestroy(&op_error); 575 CeedDestroy(&ceed); 576 return PetscFinalize(); 577 } 578