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 KSP ksp; 56 PC pc; 57 Mat *matO, *matI, *matR; 58 Vec *X, *Xloc, *mult, rhs, rhsloc, diagloc; 59 UserO *userO; 60 UserIR *userI, *userR; 61 Ceed ceed; 62 CeedData *ceeddata; 63 CeedVector rhsceed, diagceed, target; 64 CeedQFunction qf_error, qf_restrict, qf_prolong; 65 CeedOperator op_error; 66 bpType bpChoice; 67 coarsenType coarsen; 68 69 ierr = PetscInitialize(&argc, &argv, NULL, help); 70 if (ierr) return ierr; 71 comm = PETSC_COMM_WORLD; 72 73 // Parse command line options 74 ierr = PetscOptionsBegin(comm, NULL, "CEED BPs in PETSc", NULL); CHKERRQ(ierr); 75 bpChoice = CEED_BP3; 76 ierr = PetscOptionsEnum("-problem", 77 "CEED benchmark problem to solve", NULL, 78 bpTypes, (PetscEnum)bpChoice, (PetscEnum *)&bpChoice, 79 NULL); CHKERRQ(ierr); 80 ncompu = bpOptions[bpChoice].ncompu; 81 test_mode = PETSC_FALSE; 82 ierr = PetscOptionsBool("-test", 83 "Testing mode (do not print unless error is large)", 84 NULL, test_mode, &test_mode, NULL); CHKERRQ(ierr); 85 benchmark_mode = PETSC_FALSE; 86 ierr = PetscOptionsBool("-benchmark", 87 "Benchmarking mode (prints benchmark statistics)", 88 NULL, benchmark_mode, &benchmark_mode, NULL); 89 CHKERRQ(ierr); 90 write_solution = PETSC_FALSE; 91 ierr = PetscOptionsBool("-write_solution", 92 "Write solution for visualization", 93 NULL, write_solution, &write_solution, NULL); 94 CHKERRQ(ierr); 95 degree = test_mode ? 3 : 2; 96 ierr = PetscOptionsInt("-degree", "Polynomial degree of tensor product basis", 97 NULL, degree, °ree, NULL); CHKERRQ(ierr); 98 if (degree < 1) SETERRQ1(PETSC_COMM_WORLD, PETSC_ERR_ARG_OUTOFRANGE, 99 "-degree %D must be at least 1", degree); 100 qextra = bpOptions[bpChoice].qextra; 101 ierr = PetscOptionsInt("-qextra", "Number of extra quadrature points", 102 NULL, qextra, &qextra, NULL); CHKERRQ(ierr); 103 ierr = PetscOptionsString("-ceed", "CEED resource specifier", 104 NULL, ceedresource, ceedresource, 105 sizeof(ceedresource), NULL); CHKERRQ(ierr); 106 coarsen = COARSEN_UNIFORM; 107 ierr = PetscOptionsEnum("-coarsen", 108 "Coarsening strategy to use", NULL, 109 coarsenTypes, (PetscEnum)coarsen, 110 (PetscEnum *)&coarsen, NULL); CHKERRQ(ierr); 111 read_mesh = PETSC_FALSE; 112 ierr = PetscOptionsString("-mesh", "Read mesh from file", NULL, 113 filename, filename, sizeof(filename), &read_mesh); 114 CHKERRQ(ierr); 115 if (!read_mesh) { 116 PetscInt tmp = dim; 117 ierr = PetscOptionsIntArray("-cells","Number of cells per dimension", NULL, 118 melem, &tmp, NULL); CHKERRQ(ierr); 119 } 120 ierr = PetscOptionsEnd(); CHKERRQ(ierr); 121 122 // Setup DM 123 if (read_mesh) { 124 ierr = DMPlexCreateFromFile(PETSC_COMM_WORLD, filename, PETSC_TRUE, &dmOrig); 125 CHKERRQ(ierr); 126 } else { 127 ierr = DMPlexCreateBoxMesh(PETSC_COMM_WORLD, dim, PETSC_FALSE, melem, NULL, 128 NULL, NULL, PETSC_TRUE,&dmOrig); CHKERRQ(ierr); 129 } 130 131 { 132 DM dmDist = NULL; 133 PetscPartitioner part; 134 135 ierr = DMPlexGetPartitioner(dmOrig, &part); CHKERRQ(ierr); 136 ierr = PetscPartitionerSetFromOptions(part); CHKERRQ(ierr); 137 ierr = DMPlexDistribute(dmOrig, 0, NULL, &dmDist); CHKERRQ(ierr); 138 if (dmDist) { 139 ierr = DMDestroy(&dmOrig); CHKERRQ(ierr); 140 dmOrig = dmDist; 141 } 142 } 143 144 // Allocate arrays for PETSc objects for each level 145 switch (coarsen) { 146 case COARSEN_UNIFORM: 147 numlevels = degree; 148 break; 149 case COARSEN_LOGARITHMIC: 150 numlevels = ceil(log(degree)/log(2)) + 1; 151 break; 152 } 153 ierr = PetscMalloc1(numlevels, &leveldegrees); CHKERRQ(ierr); 154 switch (coarsen) { 155 case COARSEN_UNIFORM: 156 for (int i=0; i<numlevels; i++) leveldegrees[i] = i + 1; 157 break; 158 case COARSEN_LOGARITHMIC: 159 for (int i=0; i<numlevels-1; i++) leveldegrees[i] = pow(2,i); 160 leveldegrees[numlevels-1] = degree; 161 break; 162 } 163 ierr = PetscMalloc1(numlevels, &dm); CHKERRQ(ierr); 164 ierr = PetscMalloc1(numlevels, &X); CHKERRQ(ierr); 165 ierr = PetscMalloc1(numlevels, &Xloc); CHKERRQ(ierr); 166 ierr = PetscMalloc1(numlevels, &mult); CHKERRQ(ierr); 167 ierr = PetscMalloc1(numlevels, &userO); CHKERRQ(ierr); 168 ierr = PetscMalloc1(numlevels, &userI); CHKERRQ(ierr); 169 ierr = PetscMalloc1(numlevels, &userR); CHKERRQ(ierr); 170 ierr = PetscMalloc1(numlevels, &matO); CHKERRQ(ierr); 171 ierr = PetscMalloc1(numlevels, &matI); CHKERRQ(ierr); 172 ierr = PetscMalloc1(numlevels, &matR); CHKERRQ(ierr); 173 ierr = PetscMalloc1(numlevels, &lsize); CHKERRQ(ierr); 174 ierr = PetscMalloc1(numlevels, &xlsize); CHKERRQ(ierr); 175 ierr = PetscMalloc1(numlevels, &gsize); CHKERRQ(ierr); 176 177 // Setup DM and Operator Mat Shells for each level 178 for (CeedInt i=0; i<numlevels; i++) { 179 // Create DM 180 ierr = DMClone(dmOrig, &dm[i]); CHKERRQ(ierr); 181 ierr = SetupDMByDegree(dm[i], leveldegrees[i], ncompu, bpChoice); 182 CHKERRQ(ierr); 183 184 // Create vectors 185 ierr = DMCreateGlobalVector(dm[i], &X[i]); CHKERRQ(ierr); 186 ierr = VecGetLocalSize(X[i], &lsize[i]); CHKERRQ(ierr); 187 ierr = VecGetSize(X[i], &gsize[i]); CHKERRQ(ierr); 188 ierr = DMCreateLocalVector(dm[i], &Xloc[i]); CHKERRQ(ierr); 189 ierr = VecGetSize(Xloc[i], &xlsize[i]); CHKERRQ(ierr); 190 191 // Operator 192 ierr = PetscMalloc1(1, &userO[i]); CHKERRQ(ierr); 193 ierr = MatCreateShell(comm, lsize[i], lsize[i], gsize[i], gsize[i], 194 userO[i], &matO[i]); CHKERRQ(ierr); 195 ierr = MatShellSetOperation(matO[i], MATOP_MULT, 196 (void(*)(void))MatMult_Ceed); 197 ierr = MatShellSetOperation(matO[i], MATOP_GET_DIAGONAL, 198 (void(*)(void))MatGetDiag); 199 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 " Multigrid:\n" 240 " Number of Levels : %d\n", 241 bpChoice+1, usedresource, P, Q, 242 gsize[numlevels-1]/ncompu, lsize[numlevels-1]/ncompu, 243 numlevels); CHKERRQ(ierr); 244 } 245 246 // Create RHS vector 247 ierr = VecDuplicate(Xloc[numlevels-1], &rhsloc); CHKERRQ(ierr); 248 ierr = VecZeroEntries(rhsloc); CHKERRQ(ierr); 249 ierr = VecGetArray(rhsloc, &r); CHKERRQ(ierr); 250 CeedVectorCreate(ceed, xlsize[numlevels-1], &rhsceed); 251 CeedVectorSetArray(rhsceed, CEED_MEM_HOST, CEED_USE_POINTER, r); 252 253 // Set up libCEED operators on each level 254 ierr = PetscMalloc1(numlevels, &ceeddata); CHKERRQ(ierr); 255 for (int i=0; i<numlevels; i++) { 256 // Print level information 257 if (!test_mode && (i == 0 || i == numlevels-1)) { 258 ierr = PetscPrintf(comm," Level %D (%s):\n" 259 " Number of 1D Basis Nodes (p) : %d\n" 260 " Global Nodes : %D\n" 261 " Owned Nodes : %D\n", 262 i, (i? "fine" : "coarse"), leveldegrees[i] + 1, 263 gsize[i]/ncompu, lsize[i]/ncompu); CHKERRQ(ierr); 264 } 265 ierr = PetscMalloc1(1, &ceeddata[i]); CHKERRQ(ierr); 266 ierr = SetupLibceedByDegree(dm[i], ceed, leveldegrees[i], dim, qextra, 267 ncompu, gsize[i], xlsize[i], bpChoice, 268 ceeddata[i], i==(numlevels-1), rhsceed, 269 &target); CHKERRQ(ierr); 270 } 271 272 // Gather RHS 273 ierr = VecRestoreArray(rhsloc, &r); CHKERRQ(ierr); 274 ierr = VecZeroEntries(rhs); CHKERRQ(ierr); 275 ierr = DMLocalToGlobalBegin(dm[numlevels-1], rhsloc, ADD_VALUES, rhs); 276 CHKERRQ(ierr); 277 ierr = DMLocalToGlobalEnd(dm[numlevels-1], rhsloc, ADD_VALUES, rhs); 278 CHKERRQ(ierr); 279 CeedVectorDestroy(&rhsceed); 280 281 // Create the restriction/interpolation Q-function 282 CeedQFunctionCreateInterior(ceed, 1, bpOptions[bpChoice].ident, 283 bpOptions[bpChoice].identfname, &qf_restrict); 284 CeedQFunctionAddInput(qf_restrict, "uin", ncompu, CEED_EVAL_NONE); 285 CeedQFunctionAddOutput(qf_restrict, "uout", ncompu, CEED_EVAL_INTERP); 286 CeedQFunctionCreateInterior(ceed, 1, bpOptions[bpChoice].ident, 287 bpOptions[bpChoice].identfname, &qf_prolong); 288 CeedQFunctionAddInput(qf_prolong, "uin", ncompu, CEED_EVAL_INTERP); 289 CeedQFunctionAddOutput(qf_prolong, "uout", ncompu, CEED_EVAL_NONE); 290 291 // Set up libCEED level transfer operators 292 ierr = CeedLevelTransferSetup(ceed, numlevels, ncompu, bpChoice, ceeddata, 293 leveldegrees, qf_restrict, qf_prolong); 294 CHKERRQ(ierr); 295 296 // Create the error Q-function 297 CeedQFunctionCreateInterior(ceed, 1, bpOptions[bpChoice].error, 298 bpOptions[bpChoice].errorfname, &qf_error); 299 CeedQFunctionAddInput(qf_error, "u", ncompu, CEED_EVAL_INTERP); 300 CeedQFunctionAddInput(qf_error, "true_soln", ncompu, CEED_EVAL_NONE); 301 CeedQFunctionAddOutput(qf_error, "error", ncompu, CEED_EVAL_NONE); 302 303 // Create the error operator 304 CeedOperatorCreate(ceed, qf_error, NULL, NULL, &op_error); 305 CeedOperatorSetField(op_error, "u", ceeddata[numlevels-1]->Erestrictu, 306 CEED_TRANSPOSE, ceeddata[numlevels-1]->basisu, 307 CEED_VECTOR_ACTIVE); 308 CeedOperatorSetField(op_error, "true_soln", ceeddata[numlevels-1]->Erestrictui, 309 CEED_NOTRANSPOSE, CEED_BASIS_COLLOCATED, target); 310 CeedOperatorSetField(op_error, "error", ceeddata[numlevels-1]->Erestrictui, 311 CEED_NOTRANSPOSE, CEED_BASIS_COLLOCATED, 312 CEED_VECTOR_ACTIVE); 313 314 // Calculate multiplicity 315 for (int i=0; i<numlevels; i++) { 316 PetscScalar *x; 317 318 // CEED vector 319 ierr = VecGetArray(Xloc[i], &x); CHKERRQ(ierr); 320 CeedVectorSetArray(ceeddata[i]->xceed, CEED_MEM_HOST, CEED_USE_POINTER, x); 321 322 // Multiplicity 323 CeedElemRestrictionGetMultiplicity(ceeddata[i]->Erestrictu, 324 ceeddata[i]->xceed); 325 326 // Restore vector 327 ierr = VecRestoreArray(Xloc[i], &x); CHKERRQ(ierr); 328 329 // Creat mult vector 330 ierr = VecDuplicate(Xloc[i], &mult[i]); CHKERRQ(ierr); 331 332 // Local-to-global 333 ierr = VecZeroEntries(X[i]); CHKERRQ(ierr); 334 ierr = DMLocalToGlobalBegin(dm[i], Xloc[i], ADD_VALUES, X[i]); 335 CHKERRQ(ierr); 336 ierr = DMLocalToGlobalEnd(dm[i], Xloc[i], ADD_VALUES, X[i]); 337 CHKERRQ(ierr); 338 ierr = VecZeroEntries(Xloc[i]); CHKERRQ(ierr); 339 340 // Global-to-local 341 ierr = DMGlobalToLocalBegin(dm[i], X[i], INSERT_VALUES, mult[i]); 342 CHKERRQ(ierr); 343 ierr = DMGlobalToLocalEnd(dm[i], X[i], INSERT_VALUES, mult[i]); 344 CHKERRQ(ierr); 345 ierr = VecZeroEntries(X[i]); CHKERRQ(ierr); 346 347 // Multiplicity scaling 348 ierr = VecReciprocal(mult[i]); 349 } 350 351 // Set up Mat 352 for (int i=0; i<numlevels; i++) { 353 // User Operator 354 userO[i]->comm = comm; 355 userO[i]->dm = dm[i]; 356 userO[i]->Xloc = Xloc[i]; 357 ierr = VecDuplicate(Xloc[i], &userO[i]->Yloc); CHKERRQ(ierr); 358 userO[i]->xceed = ceeddata[i]->xceed; 359 userO[i]->yceed = ceeddata[i]->yceed; 360 userO[i]->op = ceeddata[i]->op_apply; 361 userO[i]->ceed = ceed; 362 363 // Set up diagonal 364 const CeedScalar *ceedarray; 365 PetscScalar *petscarray; 366 CeedInt length; 367 368 ierr = VecDuplicate(X[i], &userO[i]->diag); CHKERRQ(ierr); 369 ierr = VecDuplicate(Xloc[i], &diagloc); CHKERRQ(ierr); 370 371 // -- Local diagonal 372 CeedOperatorAssembleLinearDiagonal(userO[i]->op, &diagceed, 373 CEED_REQUEST_IMMEDIATE); 374 375 // -- Copy values 376 CeedVectorGetArrayRead(diagceed, CEED_MEM_HOST, &ceedarray); 377 ierr = VecGetArray(diagloc, &petscarray); CHKERRQ(ierr); 378 CeedVectorGetLength(diagceed, &length); 379 for (CeedInt i=0; i<length; i++) 380 petscarray[i] = ceedarray[i]; 381 CeedVectorRestoreArrayRead(diagceed, &ceedarray); 382 ierr = VecRestoreArray(diagloc, &petscarray); CHKERRQ(ierr); 383 384 // -- Global diagonal 385 ierr = VecZeroEntries(userO[i]->diag); CHKERRQ(ierr); 386 ierr = DMLocalToGlobalBegin(userO[i]->dm, diagloc, ADD_VALUES, 387 userO[i]->diag); CHKERRQ(ierr); 388 ierr = DMLocalToGlobalEnd(userO[i]->dm, diagloc, ADD_VALUES, 389 userO[i]->diag); CHKERRQ(ierr); 390 391 // -- Cleanup 392 ierr = VecDestroy(&diagloc); CHKERRQ(ierr); 393 CeedVectorDestroy(&diagceed); 394 395 if (i > 0) { 396 // Interp Operator 397 userI[i]->comm = comm; 398 userI[i]->dmc = dm[i-1]; 399 userI[i]->dmf = dm[i]; 400 userI[i]->Xloc = Xloc[i-1]; 401 userI[i]->Yloc = userO[i]->Yloc; 402 userI[i]->mult = mult[i]; 403 userI[i]->ceedvecc = userO[i-1]->xceed; 404 userI[i]->ceedvecf = userO[i]->yceed; 405 userI[i]->op = ceeddata[i]->op_interp; 406 userI[i]->ceed = ceed; 407 408 // Restrict Operator 409 userR[i]->comm = comm; 410 userR[i]->dmc = dm[i-1]; 411 userR[i]->dmf = dm[i]; 412 userR[i]->Xloc = Xloc[i]; 413 userR[i]->Yloc = userO[i-1]->Yloc; 414 userR[i]->mult = mult[i]; 415 userR[i]->ceedvecf = userO[i]->xceed; 416 userR[i]->ceedvecc = userO[i-1]->yceed; 417 userR[i]->op = ceeddata[i]->op_restrict; 418 userR[i]->ceed = ceed; 419 } 420 } 421 422 // Set up KSP 423 ierr = KSPCreate(comm, &ksp); CHKERRQ(ierr); 424 { 425 ierr = KSPSetType(ksp, KSPCG); CHKERRQ(ierr); 426 ierr = KSPSetNormType(ksp, KSP_NORM_NATURAL); CHKERRQ(ierr); 427 ierr = KSPSetTolerances(ksp, 1e-10, PETSC_DEFAULT, PETSC_DEFAULT, 428 PETSC_DEFAULT); CHKERRQ(ierr); 429 } 430 ierr = KSPSetFromOptions(ksp); CHKERRQ(ierr); 431 ierr = KSPSetOperators(ksp, matO[numlevels-1], matO[numlevels-1]); 432 CHKERRQ(ierr); 433 434 // Set up PCMG 435 ierr = KSPGetPC(ksp, &pc); CHKERRQ(ierr); 436 PCMGCycleType pcgmcycletype = PC_MG_CYCLE_V; 437 { 438 ierr = PCSetType(pc, PCMG); CHKERRQ(ierr); 439 440 // PCMG levels 441 ierr = PCMGSetLevels(pc, numlevels, NULL); CHKERRQ(ierr); 442 for (int i=0; i<numlevels; i++) { 443 // Smoother 444 KSP smoother; 445 PC smoother_pc; 446 ierr = PCMGGetSmoother(pc, i, &smoother); CHKERRQ(ierr); 447 ierr = KSPSetType(smoother, KSPCHEBYSHEV); CHKERRQ(ierr); 448 ierr = KSPChebyshevEstEigSet(smoother, 0, 0.1, 0, 1.1); CHKERRQ(ierr); 449 ierr = KSPChebyshevEstEigSetUseNoisy(smoother, PETSC_TRUE); CHKERRQ(ierr); 450 ierr = KSPSetOperators(smoother, matO[i], matO[i]); CHKERRQ(ierr); 451 ierr = KSPGetPC(smoother, &smoother_pc); CHKERRQ(ierr); 452 ierr = PCSetType(smoother_pc, PCJACOBI); CHKERRQ(ierr); 453 ierr = PCJacobiSetType(smoother_pc, PC_JACOBI_DIAGONAL); CHKERRQ(ierr); 454 455 // Work vector 456 if (i < numlevels-1) { 457 ierr = PCMGSetX(pc, i, X[i]); CHKERRQ(ierr); 458 } 459 460 // Level transfers 461 if (i > 0) { 462 // Interpolation 463 ierr = PCMGSetInterpolation(pc, i, matI[i]); CHKERRQ(ierr); 464 465 // Restriction 466 ierr = PCMGSetRestriction(pc, i, matR[i]); CHKERRQ(ierr); 467 } 468 469 // Coarse solve 470 KSP coarse; 471 PC coarse_pc; 472 ierr = PCMGGetCoarseSolve(pc, &coarse); CHKERRQ(ierr); 473 ierr = KSPSetType(coarse, KSPCG); CHKERRQ(ierr); 474 ierr = KSPSetOperators(coarse, matO[0], matO[0]); CHKERRQ(ierr); 475 ierr = KSPSetTolerances(coarse, 1e-10, 1e-10, PETSC_DEFAULT, 476 PETSC_DEFAULT); CHKERRQ(ierr); 477 ierr = KSPGetPC(coarse, &coarse_pc); CHKERRQ(ierr); 478 ierr = PCSetType(coarse_pc, PCJACOBI); CHKERRQ(ierr); 479 ierr = PCJacobiSetType(coarse_pc, PC_JACOBI_DIAGONAL); CHKERRQ(ierr); 480 } 481 482 // PCMG options 483 ierr = PCMGSetType(pc, PC_MG_MULTIPLICATIVE); CHKERRQ(ierr); 484 ierr = PCMGSetNumberSmooth(pc, 3); CHKERRQ(ierr); 485 ierr = PCMGSetCycleType(pc, pcgmcycletype); CHKERRQ(ierr); 486 } 487 488 // First run, if benchmarking 489 if (benchmark_mode) { 490 ierr = KSPSetTolerances(ksp, 1e-10, PETSC_DEFAULT, PETSC_DEFAULT, 1); 491 CHKERRQ(ierr); 492 ierr = VecZeroEntries(X[numlevels-1]); CHKERRQ(ierr); 493 my_rt_start = MPI_Wtime(); 494 ierr = KSPSolve(ksp, rhs, X[numlevels-1]); CHKERRQ(ierr); 495 my_rt = MPI_Wtime() - my_rt_start; 496 ierr = MPI_Allreduce(MPI_IN_PLACE, &my_rt, 1, MPI_DOUBLE, MPI_MIN, comm); 497 CHKERRQ(ierr); 498 // Set maxits based on first iteration timing 499 if (my_rt > 0.02) { 500 ierr = KSPSetTolerances(ksp, 1e-10, PETSC_DEFAULT, PETSC_DEFAULT, 5); 501 CHKERRQ(ierr); 502 } else { 503 ierr = KSPSetTolerances(ksp, 1e-10, PETSC_DEFAULT, PETSC_DEFAULT, 20); 504 CHKERRQ(ierr); 505 } 506 } 507 508 // Timed solve 509 ierr = VecZeroEntries(X[numlevels-1]); CHKERRQ(ierr); 510 ierr = PetscBarrier((PetscObject)ksp); CHKERRQ(ierr); 511 my_rt_start = MPI_Wtime(); 512 ierr = KSPSolve(ksp, rhs, X[numlevels-1]); CHKERRQ(ierr); 513 my_rt = MPI_Wtime() - my_rt_start; 514 515 // Output results 516 { 517 KSPType ksptype; 518 PCMGType pcmgtype; 519 KSPConvergedReason reason; 520 PetscReal rnorm; 521 PetscInt its; 522 ierr = KSPGetType(ksp, &ksptype); CHKERRQ(ierr); 523 ierr = KSPGetConvergedReason(ksp, &reason); CHKERRQ(ierr); 524 ierr = KSPGetIterationNumber(ksp, &its); CHKERRQ(ierr); 525 ierr = KSPGetResidualNorm(ksp, &rnorm); CHKERRQ(ierr); 526 ierr = PCMGGetType(pc, &pcmgtype); CHKERRQ(ierr); 527 if (!test_mode || reason < 0 || rnorm > 1e-8) { 528 ierr = PetscPrintf(comm, 529 " KSP:\n" 530 " KSP Type : %s\n" 531 " KSP Convergence : %s\n" 532 " Total KSP Iterations : %D\n" 533 " Final rnorm : %e\n", 534 ksptype, KSPConvergedReasons[reason], its, 535 (double)rnorm); CHKERRQ(ierr); 536 ierr = PetscPrintf(comm, 537 " PCMG:\n" 538 " PCMG Type : %s\n" 539 " PCMG Cycle Type : %s\n", 540 PCMGTypes[pcmgtype], 541 PCMGCycleTypes[pcgmcycletype]); CHKERRQ(ierr); 542 } 543 if (!test_mode) { 544 ierr = PetscPrintf(comm," Performance:\n"); CHKERRQ(ierr); 545 } 546 { 547 PetscReal maxerror; 548 ierr = ComputeErrorMax(userO[numlevels-1], op_error, X[numlevels-1], target, 549 &maxerror); CHKERRQ(ierr); 550 PetscReal tol = 5e-2; 551 if (!test_mode || maxerror > tol) { 552 ierr = MPI_Allreduce(&my_rt, &rt_min, 1, MPI_DOUBLE, MPI_MIN, comm); 553 CHKERRQ(ierr); 554 ierr = MPI_Allreduce(&my_rt, &rt_max, 1, MPI_DOUBLE, MPI_MAX, comm); 555 CHKERRQ(ierr); 556 ierr = PetscPrintf(comm, 557 " Pointwise Error (max) : %e\n" 558 " CG Solve Time : %g (%g) sec\n", 559 (double)maxerror, rt_max, rt_min); CHKERRQ(ierr); 560 } 561 } 562 if (benchmark_mode && (!test_mode)) { 563 ierr = PetscPrintf(comm, 564 " DoFs/Sec in CG : %g (%g) million\n", 565 1e-6*gsize[numlevels-1]*its/rt_max, 566 1e-6*gsize[numlevels-1]*its/rt_min); 567 CHKERRQ(ierr); 568 } 569 } 570 571 if (write_solution) { 572 PetscViewer vtkviewersoln; 573 574 ierr = PetscViewerCreate(comm, &vtkviewersoln); CHKERRQ(ierr); 575 ierr = PetscViewerSetType(vtkviewersoln, PETSCVIEWERVTK); CHKERRQ(ierr); 576 ierr = PetscViewerFileSetName(vtkviewersoln, "solution.vtk"); CHKERRQ(ierr); 577 ierr = VecView(X[numlevels-1], vtkviewersoln); CHKERRQ(ierr); 578 ierr = PetscViewerDestroy(&vtkviewersoln); CHKERRQ(ierr); 579 } 580 581 // Cleanup 582 for (int i=0; i<numlevels; i++) { 583 ierr = VecDestroy(&X[i]); CHKERRQ(ierr); 584 ierr = VecDestroy(&Xloc[i]); CHKERRQ(ierr); 585 ierr = VecDestroy(&mult[i]); CHKERRQ(ierr); 586 ierr = VecDestroy(&userO[i]->Yloc); CHKERRQ(ierr); 587 ierr = VecDestroy(&userO[i]->diag); CHKERRQ(ierr); 588 ierr = MatDestroy(&matO[i]); CHKERRQ(ierr); 589 ierr = PetscFree(userO[i]); CHKERRQ(ierr); 590 if (i > 0) { 591 ierr = MatDestroy(&matI[i]); CHKERRQ(ierr); 592 ierr = PetscFree(userI[i]); CHKERRQ(ierr); 593 ierr = MatDestroy(&matR[i]); CHKERRQ(ierr); 594 ierr = PetscFree(userR[i]); CHKERRQ(ierr); 595 } 596 ierr = CeedDataDestroy(i, ceeddata[i]); CHKERRQ(ierr); 597 ierr = DMDestroy(&dm[i]); CHKERRQ(ierr); 598 } 599 ierr = PetscFree(leveldegrees); CHKERRQ(ierr); 600 ierr = PetscFree(dm); CHKERRQ(ierr); 601 ierr = PetscFree(X); CHKERRQ(ierr); 602 ierr = PetscFree(Xloc); CHKERRQ(ierr); 603 ierr = PetscFree(mult); CHKERRQ(ierr); 604 ierr = PetscFree(matO); CHKERRQ(ierr); 605 ierr = PetscFree(matI); CHKERRQ(ierr); 606 ierr = PetscFree(matR); CHKERRQ(ierr); 607 ierr = PetscFree(ceeddata); CHKERRQ(ierr); 608 ierr = PetscFree(userO); CHKERRQ(ierr); 609 ierr = PetscFree(userI); CHKERRQ(ierr); 610 ierr = PetscFree(userR); CHKERRQ(ierr); 611 ierr = PetscFree(lsize); CHKERRQ(ierr); 612 ierr = PetscFree(xlsize); CHKERRQ(ierr); 613 ierr = PetscFree(gsize); CHKERRQ(ierr); 614 ierr = VecDestroy(&rhs); CHKERRQ(ierr); 615 ierr = VecDestroy(&rhsloc); CHKERRQ(ierr); 616 ierr = KSPDestroy(&ksp); CHKERRQ(ierr); 617 ierr = DMDestroy(&dmOrig); CHKERRQ(ierr); 618 CeedVectorDestroy(&target); 619 CeedQFunctionDestroy(&qf_error); 620 CeedQFunctionDestroy(&qf_restrict); 621 CeedQFunctionDestroy(&qf_prolong); 622 CeedOperatorDestroy(&op_error); 623 CeedDestroy(&ceed); 624 return PetscFinalize(); 625 } 626