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, *matI, *matR, matCoarse; 59 Vec *X, *Xloc, *mult, rhs, rhsloc; 60 UserO *userO; 61 UserIR *userI, *userR; 62 Ceed ceed; 63 CeedData *ceeddata; 64 CeedVector rhsceed, target; 65 CeedQFunction qf_error, qf_restrict, qf_prolong; 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, &userI); CHKERRQ(ierr); 170 ierr = PetscMalloc1(numlevels, &userR); CHKERRQ(ierr); 171 ierr = PetscMalloc1(numlevels, &matO); CHKERRQ(ierr); 172 ierr = PetscMalloc1(numlevels, &matI); CHKERRQ(ierr); 173 ierr = PetscMalloc1(numlevels, &matR); CHKERRQ(ierr); 174 ierr = PetscMalloc1(numlevels, &lsize); CHKERRQ(ierr); 175 ierr = PetscMalloc1(numlevels, &xlsize); CHKERRQ(ierr); 176 ierr = PetscMalloc1(numlevels, &gsize); CHKERRQ(ierr); 177 178 // Setup DM and Operator Mat Shells for each level 179 for (CeedInt i=0; i<numlevels; i++) { 180 // Create DM 181 ierr = DMClone(dmOrig, &dm[i]); CHKERRQ(ierr); 182 ierr = SetupDMByDegree(dm[i], leveldegrees[i], ncompu, bpChoice); 183 CHKERRQ(ierr); 184 185 // Create vectors 186 ierr = DMCreateGlobalVector(dm[i], &X[i]); CHKERRQ(ierr); 187 ierr = VecGetLocalSize(X[i], &lsize[i]); CHKERRQ(ierr); 188 ierr = VecGetSize(X[i], &gsize[i]); CHKERRQ(ierr); 189 ierr = DMCreateLocalVector(dm[i], &Xloc[i]); CHKERRQ(ierr); 190 ierr = VecGetSize(Xloc[i], &xlsize[i]); CHKERRQ(ierr); 191 192 // Operator 193 ierr = PetscMalloc1(1, &userO[i]); CHKERRQ(ierr); 194 ierr = MatCreateShell(comm, lsize[i], lsize[i], gsize[i], gsize[i], 195 userO[i], &matO[i]); CHKERRQ(ierr); 196 ierr = MatShellSetOperation(matO[i], MATOP_MULT, 197 (void(*)(void))MatMult_Ceed); CHKERRQ(ierr); 198 ierr = MatShellSetOperation(matO[i], MATOP_GET_DIAGONAL, 199 (void(*)(void))MatGetDiag); CHKERRQ(ierr); 200 201 // Level transfers 202 if (i > 0) { 203 // Interp 204 ierr = PetscMalloc1(1, &userI[i]); CHKERRQ(ierr); 205 ierr = MatCreateShell(comm, lsize[i], lsize[i-1], gsize[i], gsize[i-1], 206 userI[i], &matI[i]); CHKERRQ(ierr); 207 ierr = MatShellSetOperation(matI[i], MATOP_MULT, 208 (void(*)(void))MatMult_Interp); 209 CHKERRQ(ierr); 210 211 // Restrict 212 ierr = PetscMalloc1(1, &userR[i]); CHKERRQ(ierr); 213 ierr = MatCreateShell(comm, lsize[i-1], lsize[i], gsize[i-1], gsize[i], 214 userR[i], &matR[i]); CHKERRQ(ierr); 215 ierr = MatShellSetOperation(matR[i], MATOP_MULT, 216 (void(*)(void))MatMult_Restrict); 217 CHKERRQ(ierr); 218 } 219 } 220 ierr = VecDuplicate(X[numlevels-1], &rhs); CHKERRQ(ierr); 221 222 // Set up libCEED 223 CeedInit(ceedresource, &ceed); 224 225 // Print global grid information 226 if (!test_mode) { 227 PetscInt P = degree + 1, Q = P + qextra; 228 const char *usedresource; 229 CeedGetResource(ceed, &usedresource); 230 ierr = PetscPrintf(comm, 231 "\n-- CEED Benchmark Problem %d -- libCEED + PETSc + PCMG --\n" 232 " libCEED:\n" 233 " libCEED Backend : %s\n" 234 " Mesh:\n" 235 " Number of 1D Basis Nodes (p) : %d\n" 236 " Number of 1D Quadrature Points (q) : %d\n" 237 " Global Nodes : %D\n" 238 " Owned Nodes : %D\n" 239 " DoF per node : %D\n" 240 " Multigrid:\n" 241 " Number of Levels : %d\n", 242 bpChoice+1, usedresource, P, Q, 243 gsize[numlevels-1]/ncompu, lsize[numlevels-1]/ncompu, 244 ncompu, numlevels); CHKERRQ(ierr); 245 } 246 247 // Create RHS vector 248 ierr = VecDuplicate(Xloc[numlevels-1], &rhsloc); CHKERRQ(ierr); 249 ierr = VecZeroEntries(rhsloc); CHKERRQ(ierr); 250 ierr = VecGetArray(rhsloc, &r); CHKERRQ(ierr); 251 CeedVectorCreate(ceed, xlsize[numlevels-1], &rhsceed); 252 CeedVectorSetArray(rhsceed, CEED_MEM_HOST, CEED_USE_POINTER, r); 253 254 // Set up libCEED operators on each level 255 ierr = PetscMalloc1(numlevels, &ceeddata); CHKERRQ(ierr); 256 for (int i=0; i<numlevels; i++) { 257 // Print level information 258 if (!test_mode && (i == 0 || i == numlevels-1)) { 259 ierr = PetscPrintf(comm," Level %D (%s):\n" 260 " Number of 1D Basis Nodes (p) : %d\n" 261 " Global Nodes : %D\n" 262 " Owned Nodes : %D\n", 263 i, (i? "fine" : "coarse"), leveldegrees[i] + 1, 264 gsize[i]/ncompu, lsize[i]/ncompu); CHKERRQ(ierr); 265 } 266 ierr = PetscMalloc1(1, &ceeddata[i]); CHKERRQ(ierr); 267 ierr = SetupLibceedByDegree(dm[i], ceed, leveldegrees[i], dim, qextra, 268 ncompu, gsize[i], xlsize[i], bpChoice, 269 ceeddata[i], i==(numlevels-1), rhsceed, 270 &target); CHKERRQ(ierr); 271 } 272 273 // Gather RHS 274 ierr = VecRestoreArray(rhsloc, &r); CHKERRQ(ierr); 275 ierr = VecZeroEntries(rhs); CHKERRQ(ierr); 276 ierr = DMLocalToGlobal(dm[numlevels-1], rhsloc, ADD_VALUES, rhs); 277 CHKERRQ(ierr); 278 CeedVectorDestroy(&rhsceed); 279 280 // Create the restriction/interpolation Q-function 281 CeedQFunctionCreateIdentity(ceed, ncompu, CEED_EVAL_NONE, CEED_EVAL_INTERP, 282 &qf_restrict); 283 CeedQFunctionCreateIdentity(ceed, ncompu, CEED_EVAL_INTERP, CEED_EVAL_NONE, 284 &qf_prolong); 285 286 // Set up libCEED level transfer operators 287 ierr = CeedLevelTransferSetup(ceed, numlevels, ncompu, bpChoice, ceeddata, 288 leveldegrees, qf_restrict, qf_prolong); 289 CHKERRQ(ierr); 290 291 // Create the error Q-function 292 CeedQFunctionCreateInterior(ceed, 1, bpOptions[bpChoice].error, 293 bpOptions[bpChoice].errorfname, &qf_error); 294 CeedQFunctionAddInput(qf_error, "u", ncompu, CEED_EVAL_INTERP); 295 CeedQFunctionAddInput(qf_error, "true_soln", ncompu, CEED_EVAL_NONE); 296 CeedQFunctionAddOutput(qf_error, "error", ncompu, CEED_EVAL_NONE); 297 298 // Create the error operator 299 CeedOperatorCreate(ceed, qf_error, CEED_QFUNCTION_NONE, CEED_QFUNCTION_NONE, 300 &op_error); 301 CeedOperatorSetField(op_error, "u", ceeddata[numlevels-1]->Erestrictu, 302 ceeddata[numlevels-1]->basisu, CEED_VECTOR_ACTIVE); 303 CeedOperatorSetField(op_error, "true_soln", ceeddata[numlevels-1]->Erestrictui, 304 CEED_BASIS_COLLOCATED, target); 305 CeedOperatorSetField(op_error, "error", ceeddata[numlevels-1]->Erestrictui, 306 CEED_BASIS_COLLOCATED, CEED_VECTOR_ACTIVE); 307 308 // Calculate multiplicity 309 for (int i=0; i<numlevels; i++) { 310 PetscScalar *x; 311 312 // CEED vector 313 ierr = VecGetArray(Xloc[i], &x); CHKERRQ(ierr); 314 CeedVectorSetArray(ceeddata[i]->xceed, CEED_MEM_HOST, CEED_USE_POINTER, x); 315 316 // Multiplicity 317 CeedElemRestrictionGetMultiplicity(ceeddata[i]->Erestrictu, 318 ceeddata[i]->xceed); 319 320 // Restore vector 321 ierr = VecRestoreArray(Xloc[i], &x); CHKERRQ(ierr); 322 323 // Creat mult vector 324 ierr = VecDuplicate(Xloc[i], &mult[i]); CHKERRQ(ierr); 325 326 // Local-to-global 327 ierr = VecZeroEntries(X[i]); CHKERRQ(ierr); 328 ierr = DMLocalToGlobal(dm[i], Xloc[i], ADD_VALUES, X[i]); 329 CHKERRQ(ierr); 330 ierr = VecZeroEntries(Xloc[i]); CHKERRQ(ierr); 331 332 // Global-to-local 333 ierr = DMGlobalToLocal(dm[i], X[i], INSERT_VALUES, mult[i]); 334 CHKERRQ(ierr); 335 ierr = VecZeroEntries(X[i]); CHKERRQ(ierr); 336 337 // Multiplicity scaling 338 ierr = VecReciprocal(mult[i]); 339 } 340 341 // Set up Mat 342 for (int i=0; i<numlevels; i++) { 343 // User Operator 344 userO[i]->comm = comm; 345 userO[i]->dm = dm[i]; 346 userO[i]->Xloc = Xloc[i]; 347 ierr = VecDuplicate(Xloc[i], &userO[i]->Yloc); CHKERRQ(ierr); 348 userO[i]->xceed = ceeddata[i]->xceed; 349 userO[i]->yceed = ceeddata[i]->yceed; 350 userO[i]->op = ceeddata[i]->op_apply; 351 userO[i]->ceed = ceed; 352 353 if (i > 0) { 354 // Interp Operator 355 userI[i]->comm = comm; 356 userI[i]->dmc = dm[i-1]; 357 userI[i]->dmf = dm[i]; 358 userI[i]->Xloc = Xloc[i-1]; 359 userI[i]->Yloc = userO[i]->Yloc; 360 userI[i]->mult = mult[i]; 361 userI[i]->ceedvecc = userO[i-1]->xceed; 362 userI[i]->ceedvecf = userO[i]->yceed; 363 userI[i]->op = ceeddata[i]->op_interp; 364 userI[i]->ceed = ceed; 365 366 // Restrict Operator 367 userR[i]->comm = comm; 368 userR[i]->dmc = dm[i-1]; 369 userR[i]->dmf = dm[i]; 370 userR[i]->Xloc = Xloc[i]; 371 userR[i]->Yloc = userO[i-1]->Yloc; 372 userR[i]->mult = mult[i]; 373 userR[i]->ceedvecf = userO[i]->xceed; 374 userR[i]->ceedvecc = userO[i-1]->yceed; 375 userR[i]->op = ceeddata[i]->op_restrict; 376 userR[i]->ceed = ceed; 377 } 378 } 379 380 // Setup dummy SNES for AMG coarse solve 381 ierr = SNESCreate(comm, &snes_dummy); CHKERRQ(ierr); 382 ierr = SNESSetDM(snes_dummy, dm[0]); CHKERRQ(ierr); 383 ierr = SNESSetSolution(snes_dummy, X[0]); CHKERRQ(ierr); 384 385 // -- Jacobian matrix 386 ierr = DMSetMatType(dm[0], MATAIJ); CHKERRQ(ierr); 387 ierr = DMCreateMatrix(dm[0], &matCoarse); CHKERRQ(ierr); 388 ierr = SNESSetJacobian(snes_dummy, matCoarse, matCoarse, NULL, 389 NULL); CHKERRQ(ierr); 390 391 // -- Residual evaluation function 392 ierr = SNESSetFunction(snes_dummy, X[0], FormResidual_Ceed, 393 userO[0]); CHKERRQ(ierr); 394 395 // -- Form Jacobian 396 ierr = SNESComputeJacobianDefaultColor(snes_dummy, X[0], matO[0], 397 matCoarse, NULL); CHKERRQ(ierr); 398 399 // Set up KSP 400 ierr = KSPCreate(comm, &ksp); CHKERRQ(ierr); 401 { 402 ierr = KSPSetType(ksp, KSPCG); CHKERRQ(ierr); 403 ierr = KSPSetNormType(ksp, KSP_NORM_NATURAL); CHKERRQ(ierr); 404 ierr = KSPSetTolerances(ksp, 1e-10, PETSC_DEFAULT, PETSC_DEFAULT, 405 PETSC_DEFAULT); CHKERRQ(ierr); 406 } 407 ierr = KSPSetFromOptions(ksp); CHKERRQ(ierr); 408 ierr = KSPSetOperators(ksp, matO[numlevels-1], matO[numlevels-1]); 409 CHKERRQ(ierr); 410 411 // Set up PCMG 412 ierr = KSPGetPC(ksp, &pc); CHKERRQ(ierr); 413 PCMGCycleType pcgmcycletype = PC_MG_CYCLE_V; 414 { 415 ierr = PCSetType(pc, PCMG); CHKERRQ(ierr); 416 417 // PCMG levels 418 ierr = PCMGSetLevels(pc, numlevels, NULL); CHKERRQ(ierr); 419 for (int i=0; i<numlevels; i++) { 420 // Smoother 421 KSP smoother; 422 PC smoother_pc; 423 ierr = PCMGGetSmoother(pc, i, &smoother); CHKERRQ(ierr); 424 ierr = KSPSetType(smoother, KSPCHEBYSHEV); CHKERRQ(ierr); 425 ierr = KSPChebyshevEstEigSet(smoother, 0, 0.1, 0, 1.1); CHKERRQ(ierr); 426 ierr = KSPChebyshevEstEigSetUseNoisy(smoother, PETSC_TRUE); CHKERRQ(ierr); 427 ierr = KSPSetOperators(smoother, matO[i], matO[i]); CHKERRQ(ierr); 428 ierr = KSPGetPC(smoother, &smoother_pc); CHKERRQ(ierr); 429 ierr = PCSetType(smoother_pc, PCJACOBI); CHKERRQ(ierr); 430 ierr = PCJacobiSetType(smoother_pc, PC_JACOBI_DIAGONAL); CHKERRQ(ierr); 431 432 // Work vector 433 if (i < numlevels-1) { 434 ierr = PCMGSetX(pc, i, X[i]); CHKERRQ(ierr); 435 } 436 437 // Level transfers 438 if (i > 0) { 439 // Interpolation 440 ierr = PCMGSetInterpolation(pc, i, matI[i]); CHKERRQ(ierr); 441 442 // Restriction 443 ierr = PCMGSetRestriction(pc, i, matR[i]); CHKERRQ(ierr); 444 } 445 446 // Coarse solve 447 KSP coarse; 448 PC coarse_pc; 449 ierr = PCMGGetCoarseSolve(pc, &coarse); CHKERRQ(ierr); 450 ierr = KSPSetType(coarse, KSPPREONLY); CHKERRQ(ierr); 451 ierr = KSPSetOperators(coarse, matCoarse, matCoarse); CHKERRQ(ierr); 452 453 ierr = KSPGetPC(coarse, &coarse_pc); CHKERRQ(ierr); 454 ierr = PCSetType(coarse_pc, PCGAMG); CHKERRQ(ierr); 455 456 ierr = KSPSetOptionsPrefix(coarse, "coarse_"); CHKERRQ(ierr); 457 ierr = PCSetOptionsPrefix(coarse_pc, "coarse_"); CHKERRQ(ierr); 458 ierr = KSPSetFromOptions(coarse); CHKERRQ(ierr); 459 ierr = PCSetFromOptions(coarse_pc); CHKERRQ(ierr); 460 } 461 462 // PCMG options 463 ierr = PCMGSetType(pc, PC_MG_MULTIPLICATIVE); CHKERRQ(ierr); 464 ierr = PCMGSetNumberSmooth(pc, 3); CHKERRQ(ierr); 465 ierr = PCMGSetCycleType(pc, pcgmcycletype); CHKERRQ(ierr); 466 } 467 468 // First run, if benchmarking 469 if (benchmark_mode) { 470 ierr = KSPSetTolerances(ksp, 1e-10, PETSC_DEFAULT, PETSC_DEFAULT, 1); 471 CHKERRQ(ierr); 472 ierr = VecZeroEntries(X[numlevels-1]); CHKERRQ(ierr); 473 my_rt_start = MPI_Wtime(); 474 ierr = KSPSolve(ksp, rhs, X[numlevels-1]); CHKERRQ(ierr); 475 my_rt = MPI_Wtime() - my_rt_start; 476 ierr = MPI_Allreduce(MPI_IN_PLACE, &my_rt, 1, MPI_DOUBLE, MPI_MIN, comm); 477 CHKERRQ(ierr); 478 // Set maxits based on first iteration timing 479 if (my_rt > 0.02) { 480 ierr = KSPSetTolerances(ksp, 1e-10, PETSC_DEFAULT, PETSC_DEFAULT, 5); 481 CHKERRQ(ierr); 482 } else { 483 ierr = KSPSetTolerances(ksp, 1e-10, PETSC_DEFAULT, PETSC_DEFAULT, 20); 484 CHKERRQ(ierr); 485 } 486 } 487 488 // Timed solve 489 ierr = VecZeroEntries(X[numlevels-1]); CHKERRQ(ierr); 490 ierr = PetscBarrier((PetscObject)ksp); CHKERRQ(ierr); 491 my_rt_start = MPI_Wtime(); 492 ierr = KSPSolve(ksp, rhs, X[numlevels-1]); CHKERRQ(ierr); 493 my_rt = MPI_Wtime() - my_rt_start; 494 495 // Output results 496 { 497 KSPType ksptype; 498 PCMGType pcmgtype; 499 KSPConvergedReason reason; 500 PetscReal rnorm; 501 PetscInt its; 502 ierr = KSPGetType(ksp, &ksptype); CHKERRQ(ierr); 503 ierr = KSPGetConvergedReason(ksp, &reason); CHKERRQ(ierr); 504 ierr = KSPGetIterationNumber(ksp, &its); CHKERRQ(ierr); 505 ierr = KSPGetResidualNorm(ksp, &rnorm); CHKERRQ(ierr); 506 ierr = PCMGGetType(pc, &pcmgtype); CHKERRQ(ierr); 507 if (!test_mode || reason < 0 || rnorm > 1e-8) { 508 ierr = PetscPrintf(comm, 509 " KSP:\n" 510 " KSP Type : %s\n" 511 " KSP Convergence : %s\n" 512 " Total KSP Iterations : %D\n" 513 " Final rnorm : %e\n", 514 ksptype, KSPConvergedReasons[reason], its, 515 (double)rnorm); CHKERRQ(ierr); 516 ierr = PetscPrintf(comm, 517 " PCMG:\n" 518 " PCMG Type : %s\n" 519 " PCMG Cycle Type : %s\n", 520 PCMGTypes[pcmgtype], 521 PCMGCycleTypes[pcgmcycletype]); CHKERRQ(ierr); 522 } 523 if (!test_mode) { 524 ierr = PetscPrintf(comm," Performance:\n"); CHKERRQ(ierr); 525 } 526 { 527 PetscReal maxerror; 528 ierr = ComputeErrorMax(userO[numlevels-1], op_error, X[numlevels-1], target, 529 &maxerror); CHKERRQ(ierr); 530 PetscReal tol = 5e-2; 531 if (!test_mode || maxerror > tol) { 532 ierr = MPI_Allreduce(&my_rt, &rt_min, 1, MPI_DOUBLE, MPI_MIN, comm); 533 CHKERRQ(ierr); 534 ierr = MPI_Allreduce(&my_rt, &rt_max, 1, MPI_DOUBLE, MPI_MAX, comm); 535 CHKERRQ(ierr); 536 ierr = PetscPrintf(comm, 537 " Pointwise Error (max) : %e\n" 538 " CG Solve Time : %g (%g) sec\n", 539 (double)maxerror, rt_max, rt_min); CHKERRQ(ierr); 540 } 541 } 542 if (benchmark_mode && (!test_mode)) { 543 ierr = PetscPrintf(comm, 544 " DoFs/Sec in CG : %g (%g) million\n", 545 1e-6*gsize[numlevels-1]*its/rt_max, 546 1e-6*gsize[numlevels-1]*its/rt_min); 547 CHKERRQ(ierr); 548 } 549 } 550 551 if (write_solution) { 552 PetscViewer vtkviewersoln; 553 554 ierr = PetscViewerCreate(comm, &vtkviewersoln); CHKERRQ(ierr); 555 ierr = PetscViewerSetType(vtkviewersoln, PETSCVIEWERVTK); CHKERRQ(ierr); 556 ierr = PetscViewerFileSetName(vtkviewersoln, "solution.vtk"); CHKERRQ(ierr); 557 ierr = VecView(X[numlevels-1], vtkviewersoln); CHKERRQ(ierr); 558 ierr = PetscViewerDestroy(&vtkviewersoln); CHKERRQ(ierr); 559 } 560 561 // Cleanup 562 for (int i=0; i<numlevels; i++) { 563 ierr = VecDestroy(&X[i]); CHKERRQ(ierr); 564 ierr = VecDestroy(&Xloc[i]); CHKERRQ(ierr); 565 ierr = VecDestroy(&mult[i]); CHKERRQ(ierr); 566 ierr = VecDestroy(&userO[i]->Yloc); CHKERRQ(ierr); 567 ierr = MatDestroy(&matO[i]); CHKERRQ(ierr); 568 ierr = PetscFree(userO[i]); CHKERRQ(ierr); 569 if (i > 0) { 570 ierr = MatDestroy(&matI[i]); CHKERRQ(ierr); 571 ierr = PetscFree(userI[i]); CHKERRQ(ierr); 572 ierr = MatDestroy(&matR[i]); CHKERRQ(ierr); 573 ierr = PetscFree(userR[i]); CHKERRQ(ierr); 574 } 575 ierr = CeedDataDestroy(i, ceeddata[i]); CHKERRQ(ierr); 576 ierr = DMDestroy(&dm[i]); CHKERRQ(ierr); 577 } 578 ierr = PetscFree(leveldegrees); CHKERRQ(ierr); 579 ierr = PetscFree(dm); CHKERRQ(ierr); 580 ierr = PetscFree(X); CHKERRQ(ierr); 581 ierr = PetscFree(Xloc); CHKERRQ(ierr); 582 ierr = PetscFree(mult); CHKERRQ(ierr); 583 ierr = PetscFree(matO); CHKERRQ(ierr); 584 ierr = PetscFree(matI); CHKERRQ(ierr); 585 ierr = PetscFree(matR); CHKERRQ(ierr); 586 ierr = PetscFree(ceeddata); CHKERRQ(ierr); 587 ierr = PetscFree(userO); CHKERRQ(ierr); 588 ierr = PetscFree(userI); CHKERRQ(ierr); 589 ierr = PetscFree(userR); CHKERRQ(ierr); 590 ierr = PetscFree(lsize); CHKERRQ(ierr); 591 ierr = PetscFree(xlsize); CHKERRQ(ierr); 592 ierr = PetscFree(gsize); CHKERRQ(ierr); 593 ierr = VecDestroy(&rhs); CHKERRQ(ierr); 594 ierr = VecDestroy(&rhsloc); CHKERRQ(ierr); 595 ierr = MatDestroy(&matCoarse); CHKERRQ(ierr); 596 ierr = KSPDestroy(&ksp); CHKERRQ(ierr); 597 ierr = SNESDestroy(&snes_dummy); CHKERRQ(ierr); 598 ierr = DMDestroy(&dmOrig); CHKERRQ(ierr); 599 CeedVectorDestroy(&target); 600 CeedQFunctionDestroy(&qf_error); 601 CeedQFunctionDestroy(&qf_restrict); 602 CeedQFunctionDestroy(&qf_prolong); 603 CeedOperatorDestroy(&op_error); 604 CeedDestroy(&ceed); 605 return PetscFinalize(); 606 } 607