1 2 static char help[] = "Test PC redistribute on matrix with load imbalance. \n\ 3 Modified from src/ksp/ksp/tutorials/ex2.c.\n\ 4 Input parameters include:\n\ 5 -random_exact_sol : use a random exact solution vector\n\ 6 -view_exact_sol : write exact solution vector to stdout\n\ 7 -n <mesh_y> : number of mesh points\n\n"; 8 /* 9 Example: 10 mpiexec -n 8 ./ex3 -n 10000 -ksp_type cg -pc_type bjacobi -sub_pc_type icc -ksp_rtol 1.e-8 -log_view 11 mpiexec -n 8 ./ex3 -n 10000 -ksp_type preonly -pc_type redistribute -redistribute_ksp_type cg -redistribute_pc_type bjacobi -redistribute_sub_pc_type icc -redistribute_ksp_rtol 1.e-8 -log_view 12 */ 13 14 #include <petscksp.h> 15 16 int main(int argc, char **args) 17 { 18 Vec x, b, u; /* approx solution, RHS, exact solution */ 19 Mat A; /* linear system matrix */ 20 KSP ksp; /* linear solver context */ 21 PetscRandom rctx; /* random number generator context */ 22 PetscReal norm; /* norm of solution error */ 23 PetscInt i, j, Ii, J, Istart, Iend, m, n = 7, its, nloc, matdistribute = 0; 24 PetscBool flg = PETSC_FALSE; 25 PetscScalar v; 26 PetscMPIInt rank, size; 27 PetscLogStage stage; 28 29 PetscFunctionBeginUser; 30 PetscCall(PetscInitialize(&argc, &args, (char *)0, help)); 31 PetscCallMPI(MPI_Comm_size(PETSC_COMM_WORLD, &size)); 32 PetscCallMPI(MPI_Comm_rank(PETSC_COMM_WORLD, &rank)); 33 PetscCheck(size > 1, PETSC_COMM_WORLD, PETSC_ERR_WRONG_MPI_SIZE, "This example requires at least 2 MPI processes!"); 34 35 PetscCall(PetscOptionsGetInt(NULL, NULL, "-n", &n, NULL)); 36 PetscCall(PetscOptionsGetInt(NULL, NULL, "-matdistribute", &matdistribute, NULL)); 37 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 38 Compute the matrix and right-hand-side vector that define 39 the linear system, Ax = b. 40 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */ 41 switch (matdistribute) { 42 case 1: /* very imbalanced process load for matrix A */ 43 m = (1 + size) * size; 44 nloc = (rank + 1) * n; 45 if (rank == size - 1) { /* proc[size-1] stores all remaining rows */ 46 nloc = m * n; 47 for (i = 0; i < size - 1; i++) nloc -= (i + 1) * n; 48 } 49 break; 50 default: /* proc[0] and proc[1] load much smaller row blocks, the rest processes have same loads */ 51 if (rank == 0 || rank == 1) { 52 nloc = n; 53 } else { 54 nloc = 10 * n; /* 10x larger load */ 55 } 56 m = 2 + (size - 2) * 10; 57 break; 58 } 59 PetscCall(MatCreate(PETSC_COMM_WORLD, &A)); 60 PetscCall(MatSetSizes(A, nloc, nloc, PETSC_DECIDE, PETSC_DECIDE)); 61 PetscCall(MatSetFromOptions(A)); 62 PetscCall(MatMPIAIJSetPreallocation(A, 5, NULL, 5, NULL)); 63 PetscCall(MatSeqAIJSetPreallocation(A, 5, NULL)); 64 PetscCall(MatSetUp(A)); 65 66 PetscCall(MatGetOwnershipRange(A, &Istart, &Iend)); 67 nloc = Iend - Istart; 68 PetscCall(PetscSynchronizedPrintf(PETSC_COMM_WORLD, "[%d] A Istart,Iend: %" PetscInt_FMT " %" PetscInt_FMT "; nloc %" PetscInt_FMT "\n", rank, Istart, Iend, nloc)); 69 PetscCall(PetscSynchronizedFlush(PETSC_COMM_WORLD, PETSC_STDOUT)); 70 71 PetscCall(PetscLogStageRegister("Assembly", &stage)); 72 PetscCall(PetscLogStagePush(stage)); 73 for (Ii = Istart; Ii < Iend; Ii++) { 74 v = -1.0; 75 i = Ii / n; 76 j = Ii - i * n; 77 if (i > 0) { 78 J = Ii - n; 79 PetscCall(MatSetValues(A, 1, &Ii, 1, &J, &v, INSERT_VALUES)); 80 } 81 if (i < m - 1) { 82 J = Ii + n; 83 PetscCall(MatSetValues(A, 1, &Ii, 1, &J, &v, INSERT_VALUES)); 84 } 85 if (j > 0) { 86 J = Ii - 1; 87 PetscCall(MatSetValues(A, 1, &Ii, 1, &J, &v, INSERT_VALUES)); 88 } 89 if (j < n - 1) { 90 J = Ii + 1; 91 PetscCall(MatSetValues(A, 1, &Ii, 1, &J, &v, INSERT_VALUES)); 92 } 93 v = 4.0; 94 PetscCall(MatSetValues(A, 1, &Ii, 1, &Ii, &v, INSERT_VALUES)); 95 } 96 PetscCall(MatAssemblyBegin(A, MAT_FINAL_ASSEMBLY)); 97 PetscCall(MatAssemblyEnd(A, MAT_FINAL_ASSEMBLY)); 98 PetscCall(PetscLogStagePop()); 99 100 /* A is symmetric. Set symmetric flag to enable ICC/Cholesky preconditioner */ 101 PetscCall(MatSetOption(A, MAT_SYMMETRIC, PETSC_TRUE)); 102 103 /* Create parallel vectors. */ 104 PetscCall(VecCreate(PETSC_COMM_WORLD, &u)); 105 PetscCall(VecSetSizes(u, nloc, PETSC_DECIDE)); 106 PetscCall(VecSetFromOptions(u)); 107 PetscCall(VecDuplicate(u, &b)); 108 PetscCall(VecDuplicate(b, &x)); 109 110 /* Set exact solution; then compute right-hand-side vector. */ 111 PetscCall(PetscOptionsGetBool(NULL, NULL, "-random_exact_sol", &flg, NULL)); 112 if (flg) { 113 PetscCall(PetscRandomCreate(PETSC_COMM_WORLD, &rctx)); 114 PetscCall(PetscRandomSetFromOptions(rctx)); 115 PetscCall(VecSetRandom(u, rctx)); 116 PetscCall(PetscRandomDestroy(&rctx)); 117 } else { 118 PetscCall(VecSet(u, 1.0)); 119 } 120 PetscCall(MatMult(A, u, b)); 121 122 /* View the exact solution vector if desired */ 123 flg = PETSC_FALSE; 124 PetscCall(PetscOptionsGetBool(NULL, NULL, "-view_exact_sol", &flg, NULL)); 125 if (flg) PetscCall(VecView(u, PETSC_VIEWER_STDOUT_WORLD)); 126 127 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 128 Create the linear solver and set various options 129 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */ 130 PetscCall(KSPCreate(PETSC_COMM_WORLD, &ksp)); 131 PetscCall(KSPSetOperators(ksp, A, A)); 132 PetscCall(KSPSetTolerances(ksp, 1.e-2 / ((m + 1) * (n + 1)), PETSC_DEFAULT, PETSC_DEFAULT, PETSC_DEFAULT)); 133 PetscCall(KSPSetFromOptions(ksp)); 134 135 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 136 Solve the linear system 137 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */ 138 PetscCall(KSPSolve(ksp, b, x)); 139 140 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 141 Check solution and clean up 142 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */ 143 PetscCall(VecAXPY(x, -1.0, u)); 144 PetscCall(VecNorm(x, NORM_2, &norm)); 145 PetscCall(KSPGetIterationNumber(ksp, &its)); 146 PetscCall(PetscPrintf(PETSC_COMM_WORLD, "Norm of error %g iterations %" PetscInt_FMT "\n", (double)norm, its)); 147 148 /* Free work space. */ 149 PetscCall(KSPDestroy(&ksp)); 150 PetscCall(VecDestroy(&u)); 151 PetscCall(VecDestroy(&x)); 152 PetscCall(VecDestroy(&b)); 153 PetscCall(MatDestroy(&A)); 154 PetscCall(PetscFinalize()); 155 return 0; 156 } 157 158 /*TEST 159 160 test: 161 nsize: 8 162 args: -n 100 -ksp_type cg -pc_type bjacobi -sub_pc_type icc -ksp_rtol 1.e-8 163 164 test: 165 suffix: 2 166 nsize: 8 167 args: -n 100 -ksp_type preonly -pc_type redistribute -redistribute_ksp_type cg -redistribute_pc_type bjacobi -redistribute_sub_pc_type icc -redistribute_ksp_rtol 1.e-8 168 169 TEST*/ 170