1 // Copyright (c) 2017, Lawrence Livermore National Security, LLC. Produced at 2 // the Lawrence Livermore National Laboratory. LLNL-CODE-734707. All Rights 3 // reserved. See files LICENSE and NOTICE for details. 4 // 5 // This file is part of CEED, a collection of benchmarks, miniapps, software 6 // libraries and APIs for efficient high-order finite element and spectral 7 // element discretizations for exascale applications. For more information and 8 // source code availability see http://github.com/ceed. 9 // 10 // The CEED research is supported by the Exascale Computing Project 17-SC-20-SC, 11 // a collaborative effort of two U.S. Department of Energy organizations (Office 12 // of Science and the National Nuclear Security Administration) responsible for 13 // the planning and preparation of a capable exascale ecosystem, including 14 // software, applications, hardware, advanced system engineering and early 15 // testbed platforms, in support of the nation's exascale computing imperative. 16 17 // libCEED + PETSc Example: CEED BPs 3-6 with Multigrid 18 // 19 // This example demonstrates a simple usage of libCEED with PETSc to solve the 20 // CEED BP benchmark problems, see http://ceed.exascaleproject.org/bps. 21 // 22 // The code uses higher level communication protocols in DMPlex. 23 // 24 // Build with: 25 // 26 // make multigrid [PETSC_DIR=</path/to/petsc>] [CEED_DIR=</path/to/libceed>] 27 // 28 // Sample runs: 29 // 30 // multigrid -problem bp3 31 // multigrid -problem bp4 -ceed /cpu/self 32 // multigrid -problem bp5 -ceed /cpu/occa 33 // multigrid -problem bp6 -ceed /gpu/cuda 34 // 35 //TESTARGS -ceed {ceed_resource} -test -problem bp3 -degree 3 36 37 /// @file 38 /// CEED BPs 1-6 multigrid example using PETSc 39 const char help[] = "Solve CEED BPs using p-multigrid with PETSc and DMPlex\n"; 40 41 #define multigrid 42 #include "setup.h" 43 44 int main(int argc, char **argv) { 45 PetscInt ierr; 46 MPI_Comm comm; 47 char filename[PETSC_MAX_PATH_LEN], 48 ceedresource[PETSC_MAX_PATH_LEN] = "/cpu/self"; 49 double my_rt_start, my_rt, rt_min, rt_max; 50 PetscInt degree = 3, qextra, *lsize, *xlsize, *gsize, dim = 3, 51 melem[3] = {3, 3, 3}, ncompu = 1, numlevels = degree, *leveldegrees; 52 PetscScalar *r; 53 PetscBool test_mode, benchmark_mode, read_mesh, write_solution; 54 DM *dm, dmOrig; 55 SNES snes_dummy; 56 KSP ksp; 57 PC pc; 58 Mat *matO, *matPR, matCoarse; 59 Vec *X, *Xloc, *mult, rhs, rhsloc; 60 UserO *userO; 61 UserProlongRestr *userPR; 62 Ceed ceed; 63 CeedData *ceeddata; 64 CeedVector rhsceed, target; 65 CeedQFunction qf_error, qfRestrict, qfProlong; 66 CeedOperator op_error; 67 bpType bpChoice; 68 coarsenType coarsen; 69 70 ierr = PetscInitialize(&argc, &argv, NULL, help); 71 if (ierr) return ierr; 72 comm = PETSC_COMM_WORLD; 73 74 // Parse command line options 75 ierr = PetscOptionsBegin(comm, NULL, "CEED BPs in PETSc", NULL); CHKERRQ(ierr); 76 bpChoice = CEED_BP3; 77 ierr = PetscOptionsEnum("-problem", 78 "CEED benchmark problem to solve", NULL, 79 bpTypes, (PetscEnum)bpChoice, (PetscEnum *)&bpChoice, 80 NULL); CHKERRQ(ierr); 81 ncompu = bpOptions[bpChoice].ncompu; 82 test_mode = PETSC_FALSE; 83 ierr = PetscOptionsBool("-test", 84 "Testing mode (do not print unless error is large)", 85 NULL, test_mode, &test_mode, NULL); CHKERRQ(ierr); 86 benchmark_mode = PETSC_FALSE; 87 ierr = PetscOptionsBool("-benchmark", 88 "Benchmarking mode (prints benchmark statistics)", 89 NULL, benchmark_mode, &benchmark_mode, NULL); 90 CHKERRQ(ierr); 91 write_solution = PETSC_FALSE; 92 ierr = PetscOptionsBool("-write_solution", 93 "Write solution for visualization", 94 NULL, write_solution, &write_solution, NULL); 95 CHKERRQ(ierr); 96 degree = test_mode ? 3 : 2; 97 ierr = PetscOptionsInt("-degree", "Polynomial degree of tensor product basis", 98 NULL, degree, °ree, NULL); CHKERRQ(ierr); 99 if (degree < 1) SETERRQ1(PETSC_COMM_WORLD, PETSC_ERR_ARG_OUTOFRANGE, 100 "-degree %D must be at least 1", degree); 101 qextra = bpOptions[bpChoice].qextra; 102 ierr = PetscOptionsInt("-qextra", "Number of extra quadrature points", 103 NULL, qextra, &qextra, NULL); CHKERRQ(ierr); 104 ierr = PetscOptionsString("-ceed", "CEED resource specifier", 105 NULL, ceedresource, ceedresource, 106 sizeof(ceedresource), NULL); CHKERRQ(ierr); 107 coarsen = COARSEN_UNIFORM; 108 ierr = PetscOptionsEnum("-coarsen", 109 "Coarsening strategy to use", NULL, 110 coarsenTypes, (PetscEnum)coarsen, 111 (PetscEnum *)&coarsen, NULL); CHKERRQ(ierr); 112 read_mesh = PETSC_FALSE; 113 ierr = PetscOptionsString("-mesh", "Read mesh from file", NULL, 114 filename, filename, sizeof(filename), &read_mesh); 115 CHKERRQ(ierr); 116 if (!read_mesh) { 117 PetscInt tmp = dim; 118 ierr = PetscOptionsIntArray("-cells","Number of cells per dimension", NULL, 119 melem, &tmp, NULL); CHKERRQ(ierr); 120 } 121 ierr = PetscOptionsEnd(); CHKERRQ(ierr); 122 123 // Setup DM 124 if (read_mesh) { 125 ierr = DMPlexCreateFromFile(PETSC_COMM_WORLD, filename, PETSC_TRUE, &dmOrig); 126 CHKERRQ(ierr); 127 } else { 128 ierr = DMPlexCreateBoxMesh(PETSC_COMM_WORLD, dim, PETSC_FALSE, melem, NULL, 129 NULL, NULL, PETSC_TRUE,&dmOrig); CHKERRQ(ierr); 130 } 131 132 { 133 DM dmDist = NULL; 134 PetscPartitioner part; 135 136 ierr = DMPlexGetPartitioner(dmOrig, &part); CHKERRQ(ierr); 137 ierr = PetscPartitionerSetFromOptions(part); CHKERRQ(ierr); 138 ierr = DMPlexDistribute(dmOrig, 0, NULL, &dmDist); CHKERRQ(ierr); 139 if (dmDist) { 140 ierr = DMDestroy(&dmOrig); CHKERRQ(ierr); 141 dmOrig = dmDist; 142 } 143 } 144 145 // Allocate arrays for PETSc objects for each level 146 switch (coarsen) { 147 case COARSEN_UNIFORM: 148 numlevels = degree; 149 break; 150 case COARSEN_LOGARITHMIC: 151 numlevels = ceil(log(degree)/log(2)) + 1; 152 break; 153 } 154 ierr = PetscMalloc1(numlevels, &leveldegrees); CHKERRQ(ierr); 155 switch (coarsen) { 156 case COARSEN_UNIFORM: 157 for (int i=0; i<numlevels; i++) leveldegrees[i] = i + 1; 158 break; 159 case COARSEN_LOGARITHMIC: 160 for (int i=0; i<numlevels-1; i++) leveldegrees[i] = pow(2,i); 161 leveldegrees[numlevels-1] = degree; 162 break; 163 } 164 ierr = PetscMalloc1(numlevels, &dm); CHKERRQ(ierr); 165 ierr = PetscMalloc1(numlevels, &X); CHKERRQ(ierr); 166 ierr = PetscMalloc1(numlevels, &Xloc); CHKERRQ(ierr); 167 ierr = PetscMalloc1(numlevels, &mult); CHKERRQ(ierr); 168 ierr = PetscMalloc1(numlevels, &userO); CHKERRQ(ierr); 169 ierr = PetscMalloc1(numlevels, &userPR); CHKERRQ(ierr); 170 ierr = PetscMalloc1(numlevels, &matO); CHKERRQ(ierr); 171 ierr = PetscMalloc1(numlevels, &matPR); CHKERRQ(ierr); 172 ierr = PetscMalloc1(numlevels, &lsize); CHKERRQ(ierr); 173 ierr = PetscMalloc1(numlevels, &xlsize); CHKERRQ(ierr); 174 ierr = PetscMalloc1(numlevels, &gsize); CHKERRQ(ierr); 175 176 // Setup DM and Operator Mat Shells for each level 177 for (CeedInt i=0; i<numlevels; i++) { 178 // Create DM 179 ierr = DMClone(dmOrig, &dm[i]); CHKERRQ(ierr); 180 ierr = SetupDMByDegree(dm[i], leveldegrees[i], ncompu, bpChoice); 181 CHKERRQ(ierr); 182 183 // Create vectors 184 ierr = DMCreateGlobalVector(dm[i], &X[i]); CHKERRQ(ierr); 185 ierr = VecGetLocalSize(X[i], &lsize[i]); CHKERRQ(ierr); 186 ierr = VecGetSize(X[i], &gsize[i]); CHKERRQ(ierr); 187 ierr = DMCreateLocalVector(dm[i], &Xloc[i]); CHKERRQ(ierr); 188 ierr = VecGetSize(Xloc[i], &xlsize[i]); CHKERRQ(ierr); 189 190 // Operator 191 ierr = PetscMalloc1(1, &userO[i]); CHKERRQ(ierr); 192 ierr = MatCreateShell(comm, lsize[i], lsize[i], gsize[i], gsize[i], 193 userO[i], &matO[i]); CHKERRQ(ierr); 194 ierr = MatShellSetOperation(matO[i], MATOP_MULT, 195 (void(*)(void))MatMult_Ceed); CHKERRQ(ierr); 196 ierr = MatShellSetOperation(matO[i], MATOP_GET_DIAGONAL, 197 (void(*)(void))MatGetDiag); CHKERRQ(ierr); 198 199 // Level transfers 200 if (i > 0) { 201 // Interp 202 ierr = PetscMalloc1(1, &userPR[i]); CHKERRQ(ierr); 203 ierr = MatCreateShell(comm, lsize[i], lsize[i-1], gsize[i], gsize[i-1], 204 userPR[i], &matPR[i]); CHKERRQ(ierr); 205 ierr = MatShellSetOperation(matPR[i], MATOP_MULT, 206 (void(*)(void))MatMult_Prolong); 207 CHKERRQ(ierr); 208 ierr = MatShellSetOperation(matPR[i], MATOP_MULT_TRANSPOSE, 209 (void(*)(void))MatMult_Restrict); 210 CHKERRQ(ierr); 211 } 212 } 213 ierr = VecDuplicate(X[numlevels-1], &rhs); CHKERRQ(ierr); 214 215 // Set up libCEED 216 CeedInit(ceedresource, &ceed); 217 218 // Print global grid information 219 if (!test_mode) { 220 PetscInt P = degree + 1, Q = P + qextra; 221 const char *usedresource; 222 CeedGetResource(ceed, &usedresource); 223 ierr = PetscPrintf(comm, 224 "\n-- CEED Benchmark Problem %d -- libCEED + PETSc + PCMG --\n" 225 " libCEED:\n" 226 " libCEED Backend : %s\n" 227 " Mesh:\n" 228 " Number of 1D Basis Nodes (p) : %d\n" 229 " Number of 1D Quadrature Points (q) : %d\n" 230 " Global Nodes : %D\n" 231 " Owned Nodes : %D\n" 232 " DoF per node : %D\n" 233 " Multigrid:\n" 234 " Number of Levels : %d\n", 235 bpChoice+1, usedresource, P, Q, 236 gsize[numlevels-1]/ncompu, lsize[numlevels-1]/ncompu, 237 ncompu, numlevels); CHKERRQ(ierr); 238 } 239 240 // Create RHS vector 241 ierr = VecDuplicate(Xloc[numlevels-1], &rhsloc); CHKERRQ(ierr); 242 ierr = VecZeroEntries(rhsloc); CHKERRQ(ierr); 243 ierr = VecGetArray(rhsloc, &r); CHKERRQ(ierr); 244 CeedVectorCreate(ceed, xlsize[numlevels-1], &rhsceed); 245 CeedVectorSetArray(rhsceed, CEED_MEM_HOST, CEED_USE_POINTER, r); 246 247 // Set up libCEED operators on each level 248 ierr = PetscMalloc1(numlevels, &ceeddata); CHKERRQ(ierr); 249 for (int i=0; i<numlevels; i++) { 250 // Print level information 251 if (!test_mode && (i == 0 || i == numlevels-1)) { 252 ierr = PetscPrintf(comm," Level %D (%s):\n" 253 " Number of 1D Basis Nodes (p) : %d\n" 254 " Global Nodes : %D\n" 255 " Owned Nodes : %D\n", 256 i, (i? "fine" : "coarse"), leveldegrees[i] + 1, 257 gsize[i]/ncompu, lsize[i]/ncompu); CHKERRQ(ierr); 258 } 259 ierr = PetscMalloc1(1, &ceeddata[i]); CHKERRQ(ierr); 260 ierr = SetupLibceedByDegree(dm[i], ceed, leveldegrees[i], dim, qextra, 261 ncompu, gsize[i], xlsize[i], bpChoice, 262 ceeddata[i], i==(numlevels-1), rhsceed, 263 &target); CHKERRQ(ierr); 264 } 265 266 // Gather RHS 267 ierr = VecRestoreArray(rhsloc, &r); CHKERRQ(ierr); 268 ierr = VecZeroEntries(rhs); CHKERRQ(ierr); 269 ierr = DMLocalToGlobal(dm[numlevels-1], rhsloc, ADD_VALUES, rhs); 270 CHKERRQ(ierr); 271 CeedVectorDestroy(&rhsceed); 272 273 // Create the restriction/interpolation Q-function 274 CeedQFunctionCreateIdentity(ceed, ncompu, CEED_EVAL_NONE, CEED_EVAL_INTERP, 275 &qfRestrict); 276 CeedQFunctionCreateIdentity(ceed, ncompu, CEED_EVAL_INTERP, CEED_EVAL_NONE, 277 &qfProlong); 278 279 // Set up libCEED level transfer operators 280 ierr = CeedLevelTransferSetup(ceed, numlevels, ncompu, bpChoice, ceeddata, 281 leveldegrees, qfRestrict, qfProlong); 282 CHKERRQ(ierr); 283 284 // Create the error Q-function 285 CeedQFunctionCreateInterior(ceed, 1, bpOptions[bpChoice].error, 286 bpOptions[bpChoice].errorfname, &qf_error); 287 CeedQFunctionAddInput(qf_error, "u", ncompu, CEED_EVAL_INTERP); 288 CeedQFunctionAddInput(qf_error, "true_soln", ncompu, CEED_EVAL_NONE); 289 CeedQFunctionAddOutput(qf_error, "error", ncompu, CEED_EVAL_NONE); 290 291 // Create the error operator 292 CeedOperatorCreate(ceed, qf_error, CEED_QFUNCTION_NONE, CEED_QFUNCTION_NONE, 293 &op_error); 294 CeedOperatorSetField(op_error, "u", ceeddata[numlevels-1]->Erestrictu, 295 ceeddata[numlevels-1]->basisu, CEED_VECTOR_ACTIVE); 296 CeedOperatorSetField(op_error, "true_soln", ceeddata[numlevels-1]->Erestrictui, 297 CEED_BASIS_COLLOCATED, target); 298 CeedOperatorSetField(op_error, "error", ceeddata[numlevels-1]->Erestrictui, 299 CEED_BASIS_COLLOCATED, CEED_VECTOR_ACTIVE); 300 301 // Calculate multiplicity 302 for (int i=0; i<numlevels; i++) { 303 PetscScalar *x; 304 305 // CEED vector 306 ierr = VecGetArray(Xloc[i], &x); CHKERRQ(ierr); 307 CeedVectorSetArray(ceeddata[i]->xceed, CEED_MEM_HOST, CEED_USE_POINTER, x); 308 309 // Multiplicity 310 CeedElemRestrictionGetMultiplicity(ceeddata[i]->Erestrictu, 311 ceeddata[i]->xceed); 312 313 // Restore vector 314 ierr = VecRestoreArray(Xloc[i], &x); CHKERRQ(ierr); 315 316 // Creat mult vector 317 ierr = VecDuplicate(Xloc[i], &mult[i]); CHKERRQ(ierr); 318 319 // Local-to-global 320 ierr = VecZeroEntries(X[i]); CHKERRQ(ierr); 321 ierr = DMLocalToGlobal(dm[i], Xloc[i], ADD_VALUES, X[i]); 322 CHKERRQ(ierr); 323 ierr = VecZeroEntries(Xloc[i]); CHKERRQ(ierr); 324 325 // Global-to-local 326 ierr = DMGlobalToLocal(dm[i], X[i], INSERT_VALUES, mult[i]); 327 CHKERRQ(ierr); 328 ierr = VecZeroEntries(X[i]); CHKERRQ(ierr); 329 330 // Multiplicity scaling 331 ierr = VecReciprocal(mult[i]); 332 } 333 334 // Set up Mat 335 for (int i=0; i<numlevels; i++) { 336 // User Operator 337 userO[i]->comm = comm; 338 userO[i]->dm = dm[i]; 339 userO[i]->Xloc = Xloc[i]; 340 ierr = VecDuplicate(Xloc[i], &userO[i]->Yloc); CHKERRQ(ierr); 341 userO[i]->xceed = ceeddata[i]->xceed; 342 userO[i]->yceed = ceeddata[i]->yceed; 343 userO[i]->op = ceeddata[i]->op_apply; 344 userO[i]->ceed = ceed; 345 346 if (i > 0) { 347 // Prolongation/Restriction Operator 348 userPR[i]->comm = comm; 349 userPR[i]->dmF = dm[i]; 350 userPR[i]->dmC = dm[i-1]; 351 userPR[i]->locVecC = Xloc[i-1]; 352 userPR[i]->locVecF = userO[i]->Yloc; 353 userPR[i]->multVec = mult[i]; 354 userPR[i]->ceedVecC = userO[i-1]->xceed; 355 userPR[i]->ceedVecF = userO[i]->yceed; 356 userPR[i]->opProlong = ceeddata[i]->opProlong; 357 userPR[i]->opRestrict = ceeddata[i]->opRestrict; 358 userPR[i]->ceed = ceed; 359 } 360 } 361 362 // Setup dummy SNES for AMG coarse solve 363 ierr = SNESCreate(comm, &snes_dummy); CHKERRQ(ierr); 364 ierr = SNESSetDM(snes_dummy, dm[0]); CHKERRQ(ierr); 365 ierr = SNESSetSolution(snes_dummy, X[0]); CHKERRQ(ierr); 366 367 // -- Jacobian matrix 368 ierr = DMSetMatType(dm[0], MATAIJ); CHKERRQ(ierr); 369 ierr = DMCreateMatrix(dm[0], &matCoarse); CHKERRQ(ierr); 370 ierr = SNESSetJacobian(snes_dummy, matCoarse, matCoarse, NULL, 371 NULL); CHKERRQ(ierr); 372 373 // -- Residual evaluation function 374 ierr = SNESSetFunction(snes_dummy, X[0], FormResidual_Ceed, 375 userO[0]); CHKERRQ(ierr); 376 377 // -- Form Jacobian 378 ierr = SNESComputeJacobianDefaultColor(snes_dummy, X[0], matO[0], 379 matCoarse, NULL); CHKERRQ(ierr); 380 381 // Set up KSP 382 ierr = KSPCreate(comm, &ksp); CHKERRQ(ierr); 383 { 384 ierr = KSPSetType(ksp, KSPCG); CHKERRQ(ierr); 385 ierr = KSPSetNormType(ksp, KSP_NORM_NATURAL); CHKERRQ(ierr); 386 ierr = KSPSetTolerances(ksp, 1e-10, PETSC_DEFAULT, PETSC_DEFAULT, 387 PETSC_DEFAULT); CHKERRQ(ierr); 388 } 389 ierr = KSPSetFromOptions(ksp); CHKERRQ(ierr); 390 ierr = KSPSetOperators(ksp, matO[numlevels-1], matO[numlevels-1]); 391 CHKERRQ(ierr); 392 393 // Set up PCMG 394 ierr = KSPGetPC(ksp, &pc); CHKERRQ(ierr); 395 PCMGCycleType pcgmcycletype = PC_MG_CYCLE_V; 396 { 397 ierr = PCSetType(pc, PCMG); CHKERRQ(ierr); 398 399 // PCMG levels 400 ierr = PCMGSetLevels(pc, numlevels, NULL); CHKERRQ(ierr); 401 for (int i=0; i<numlevels; i++) { 402 // Smoother 403 KSP smoother; 404 PC smoother_pc; 405 ierr = PCMGGetSmoother(pc, i, &smoother); CHKERRQ(ierr); 406 ierr = KSPSetType(smoother, KSPCHEBYSHEV); CHKERRQ(ierr); 407 ierr = KSPChebyshevEstEigSet(smoother, 0, 0.1, 0, 1.1); CHKERRQ(ierr); 408 ierr = KSPChebyshevEstEigSetUseNoisy(smoother, PETSC_TRUE); CHKERRQ(ierr); 409 ierr = KSPSetOperators(smoother, matO[i], matO[i]); CHKERRQ(ierr); 410 ierr = KSPGetPC(smoother, &smoother_pc); CHKERRQ(ierr); 411 ierr = PCSetType(smoother_pc, PCJACOBI); CHKERRQ(ierr); 412 ierr = PCJacobiSetType(smoother_pc, PC_JACOBI_DIAGONAL); CHKERRQ(ierr); 413 414 // Work vector 415 if (i < numlevels-1) { 416 ierr = PCMGSetX(pc, i, X[i]); CHKERRQ(ierr); 417 } 418 419 // Level transfers 420 if (i > 0) { 421 // Interpolation 422 ierr = PCMGSetInterpolation(pc, i, matPR[i]); CHKERRQ(ierr); 423 } 424 425 // Coarse solve 426 KSP coarse; 427 PC coarse_pc; 428 ierr = PCMGGetCoarseSolve(pc, &coarse); CHKERRQ(ierr); 429 ierr = KSPSetType(coarse, KSPPREONLY); CHKERRQ(ierr); 430 ierr = KSPSetOperators(coarse, matCoarse, matCoarse); CHKERRQ(ierr); 431 432 ierr = KSPGetPC(coarse, &coarse_pc); CHKERRQ(ierr); 433 ierr = PCSetType(coarse_pc, PCGAMG); CHKERRQ(ierr); 434 435 ierr = KSPSetOptionsPrefix(coarse, "coarse_"); CHKERRQ(ierr); 436 ierr = PCSetOptionsPrefix(coarse_pc, "coarse_"); CHKERRQ(ierr); 437 ierr = KSPSetFromOptions(coarse); CHKERRQ(ierr); 438 ierr = PCSetFromOptions(coarse_pc); CHKERRQ(ierr); 439 } 440 441 // PCMG options 442 ierr = PCMGSetType(pc, PC_MG_MULTIPLICATIVE); CHKERRQ(ierr); 443 ierr = PCMGSetNumberSmooth(pc, 3); CHKERRQ(ierr); 444 ierr = PCMGSetCycleType(pc, pcgmcycletype); CHKERRQ(ierr); 445 } 446 447 // First run, if benchmarking 448 if (benchmark_mode) { 449 ierr = KSPSetTolerances(ksp, 1e-10, PETSC_DEFAULT, PETSC_DEFAULT, 1); 450 CHKERRQ(ierr); 451 ierr = VecZeroEntries(X[numlevels-1]); CHKERRQ(ierr); 452 my_rt_start = MPI_Wtime(); 453 ierr = KSPSolve(ksp, rhs, X[numlevels-1]); CHKERRQ(ierr); 454 my_rt = MPI_Wtime() - my_rt_start; 455 ierr = MPI_Allreduce(MPI_IN_PLACE, &my_rt, 1, MPI_DOUBLE, MPI_MIN, comm); 456 CHKERRQ(ierr); 457 // Set maxits based on first iteration timing 458 if (my_rt > 0.02) { 459 ierr = KSPSetTolerances(ksp, 1e-10, PETSC_DEFAULT, PETSC_DEFAULT, 5); 460 CHKERRQ(ierr); 461 } else { 462 ierr = KSPSetTolerances(ksp, 1e-10, PETSC_DEFAULT, PETSC_DEFAULT, 20); 463 CHKERRQ(ierr); 464 } 465 } 466 467 // Timed solve 468 ierr = VecZeroEntries(X[numlevels-1]); CHKERRQ(ierr); 469 ierr = PetscBarrier((PetscObject)ksp); CHKERRQ(ierr); 470 my_rt_start = MPI_Wtime(); 471 ierr = KSPSolve(ksp, rhs, X[numlevels-1]); CHKERRQ(ierr); 472 my_rt = MPI_Wtime() - my_rt_start; 473 474 // Output results 475 { 476 KSPType ksptype; 477 PCMGType pcmgtype; 478 KSPConvergedReason reason; 479 PetscReal rnorm; 480 PetscInt its; 481 ierr = KSPGetType(ksp, &ksptype); CHKERRQ(ierr); 482 ierr = KSPGetConvergedReason(ksp, &reason); CHKERRQ(ierr); 483 ierr = KSPGetIterationNumber(ksp, &its); CHKERRQ(ierr); 484 ierr = KSPGetResidualNorm(ksp, &rnorm); CHKERRQ(ierr); 485 ierr = PCMGGetType(pc, &pcmgtype); CHKERRQ(ierr); 486 if (!test_mode || reason < 0 || rnorm > 1e-8) { 487 ierr = PetscPrintf(comm, 488 " KSP:\n" 489 " KSP Type : %s\n" 490 " KSP Convergence : %s\n" 491 " Total KSP Iterations : %D\n" 492 " Final rnorm : %e\n", 493 ksptype, KSPConvergedReasons[reason], its, 494 (double)rnorm); CHKERRQ(ierr); 495 ierr = PetscPrintf(comm, 496 " PCMG:\n" 497 " PCMG Type : %s\n" 498 " PCMG Cycle Type : %s\n", 499 PCMGTypes[pcmgtype], 500 PCMGCycleTypes[pcgmcycletype]); CHKERRQ(ierr); 501 } 502 if (!test_mode) { 503 ierr = PetscPrintf(comm," Performance:\n"); CHKERRQ(ierr); 504 } 505 { 506 PetscReal maxerror; 507 ierr = ComputeErrorMax(userO[numlevels-1], op_error, X[numlevels-1], target, 508 &maxerror); CHKERRQ(ierr); 509 PetscReal tol = 5e-2; 510 if (!test_mode || maxerror > tol) { 511 ierr = MPI_Allreduce(&my_rt, &rt_min, 1, MPI_DOUBLE, MPI_MIN, comm); 512 CHKERRQ(ierr); 513 ierr = MPI_Allreduce(&my_rt, &rt_max, 1, MPI_DOUBLE, MPI_MAX, comm); 514 CHKERRQ(ierr); 515 ierr = PetscPrintf(comm, 516 " Pointwise Error (max) : %e\n" 517 " CG Solve Time : %g (%g) sec\n", 518 (double)maxerror, rt_max, rt_min); CHKERRQ(ierr); 519 } 520 } 521 if (benchmark_mode && (!test_mode)) { 522 ierr = PetscPrintf(comm, 523 " DoFs/Sec in CG : %g (%g) million\n", 524 1e-6*gsize[numlevels-1]*its/rt_max, 525 1e-6*gsize[numlevels-1]*its/rt_min); 526 CHKERRQ(ierr); 527 } 528 } 529 530 if (write_solution) { 531 PetscViewer vtkviewersoln; 532 533 ierr = PetscViewerCreate(comm, &vtkviewersoln); CHKERRQ(ierr); 534 ierr = PetscViewerSetType(vtkviewersoln, PETSCVIEWERVTK); CHKERRQ(ierr); 535 ierr = PetscViewerFileSetName(vtkviewersoln, "solution.vtk"); CHKERRQ(ierr); 536 ierr = VecView(X[numlevels-1], vtkviewersoln); CHKERRQ(ierr); 537 ierr = PetscViewerDestroy(&vtkviewersoln); CHKERRQ(ierr); 538 } 539 540 // Cleanup 541 for (int i=0; i<numlevels; i++) { 542 ierr = VecDestroy(&X[i]); CHKERRQ(ierr); 543 ierr = VecDestroy(&Xloc[i]); CHKERRQ(ierr); 544 ierr = VecDestroy(&mult[i]); CHKERRQ(ierr); 545 ierr = VecDestroy(&userO[i]->Yloc); CHKERRQ(ierr); 546 ierr = MatDestroy(&matO[i]); CHKERRQ(ierr); 547 ierr = PetscFree(userO[i]); CHKERRQ(ierr); 548 if (i > 0) { 549 ierr = MatDestroy(&matPR[i]); CHKERRQ(ierr); 550 ierr = PetscFree(userPR[i]); CHKERRQ(ierr); 551 } 552 ierr = CeedDataDestroy(i, ceeddata[i]); CHKERRQ(ierr); 553 ierr = DMDestroy(&dm[i]); CHKERRQ(ierr); 554 } 555 ierr = PetscFree(leveldegrees); CHKERRQ(ierr); 556 ierr = PetscFree(dm); CHKERRQ(ierr); 557 ierr = PetscFree(X); CHKERRQ(ierr); 558 ierr = PetscFree(Xloc); CHKERRQ(ierr); 559 ierr = PetscFree(mult); CHKERRQ(ierr); 560 ierr = PetscFree(matO); CHKERRQ(ierr); 561 ierr = PetscFree(matPR); CHKERRQ(ierr); 562 ierr = PetscFree(ceeddata); CHKERRQ(ierr); 563 ierr = PetscFree(userO); CHKERRQ(ierr); 564 ierr = PetscFree(userPR); CHKERRQ(ierr); 565 ierr = PetscFree(lsize); CHKERRQ(ierr); 566 ierr = PetscFree(xlsize); CHKERRQ(ierr); 567 ierr = PetscFree(gsize); CHKERRQ(ierr); 568 ierr = VecDestroy(&rhs); CHKERRQ(ierr); 569 ierr = VecDestroy(&rhsloc); CHKERRQ(ierr); 570 ierr = MatDestroy(&matCoarse); CHKERRQ(ierr); 571 ierr = KSPDestroy(&ksp); CHKERRQ(ierr); 572 ierr = SNESDestroy(&snes_dummy); CHKERRQ(ierr); 573 ierr = DMDestroy(&dmOrig); CHKERRQ(ierr); 574 CeedVectorDestroy(&target); 575 CeedQFunctionDestroy(&qf_error); 576 CeedQFunctionDestroy(&qfRestrict); 577 CeedQFunctionDestroy(&qfProlong); 578 CeedOperatorDestroy(&op_error); 579 CeedDestroy(&ceed); 580 return PetscFinalize(); 581 } 582