Actual source code: ex32.c

  1: /*
  2: Laplacian in 2D. Modeled by the partial differential equation

  4:    div  grad u = f,  0 < x,y < 1,

  6: with forcing function

  8:    f = e^{-(1 - x)^2/\nu} e^{-(1 - y)^2/\nu}

 10: with pure Neumann boundary conditions

 12: The functions are cell-centered

 14: This uses multigrid to solve the linear system

 16:        Contributed by Andrei Draganescu <aidraga@sandia.gov>

 18: Note the nice multigrid convergence despite the fact it is only using
 19: piecewise constant interpolation/restriction. This is because cell-centered multigrid
 20: does not need the same rule:

 22:     polynomial degree(interpolation) + polynomial degree(restriction) + 2 > degree of PDE

 24: that vertex based multigrid needs.
 25: */

 27: static char help[] = "Solves 2D inhomogeneous Laplacian using multigrid.\n\n";

 29: #include <petscdm.h>
 30: #include <petscdmda.h>
 31: #include <petscksp.h>

 33: extern PetscErrorCode ComputeMatrix(KSP, Mat, Mat, void *);
 34: extern PetscErrorCode ComputeRHS(KSP, Vec, void *);

 36: typedef enum {
 37:   DIRICHLET,
 38:   NEUMANN
 39: } BCType;

 41: typedef struct {
 42:   PetscScalar nu;
 43:   BCType      bcType;
 44: } UserContext;

 46: int main(int argc, char **argv)
 47: {
 48:   KSP         ksp;
 49:   DM          da;
 50:   UserContext user;
 51:   const char *bcTypes[2] = {"dirichlet", "neumann"};
 52:   PetscInt    bc;

 54:   PetscFunctionBeginUser;
 55:   PetscCall(PetscInitialize(&argc, &argv, NULL, help));
 56:   PetscCall(KSPCreate(PETSC_COMM_WORLD, &ksp));
 57:   PetscCall(DMDACreate2d(PETSC_COMM_WORLD, DM_BOUNDARY_NONE, DM_BOUNDARY_NONE, DMDA_STENCIL_STAR, 2, 2, PETSC_DECIDE, PETSC_DECIDE, 1, 1, 0, 0, &da));
 58:   PetscCall(DMDASetInterpolationType(da, DMDA_Q0));
 59:   PetscCall(DMSetFromOptions(da));
 60:   PetscCall(DMSetUp(da));

 62:   PetscCall(KSPSetDM(ksp, da));

 64:   PetscOptionsBegin(PETSC_COMM_WORLD, "", "Options for the inhomogeneous Poisson equation", "DM");
 65:   user.nu = 0.1;
 66:   PetscCall(PetscOptionsScalar("-nu", "The width of the Gaussian source", "ex29.c", 0.1, &user.nu, NULL));
 67:   bc = (PetscInt)NEUMANN;
 68:   PetscCall(PetscOptionsEList("-bc_type", "Type of boundary condition", "ex29.c", bcTypes, 2, bcTypes[0], &bc, NULL));
 69:   user.bcType = (BCType)bc;
 70:   PetscOptionsEnd();

 72:   PetscCall(KSPSetComputeRHS(ksp, ComputeRHS, &user));
 73:   PetscCall(KSPSetComputeOperators(ksp, ComputeMatrix, &user));
 74:   PetscCall(KSPSetFromOptions(ksp));
 75:   PetscCall(KSPSetUp(ksp));
 76:   PetscCall(KSPSolve(ksp, NULL, NULL));
 77:   PetscCall(KSPDestroy(&ksp));
 78:   PetscCall(DMDestroy(&da));
 79:   PetscCall(PetscFinalize());
 80:   return 0;
 81: }

 83: PetscErrorCode ComputeRHS(KSP ksp, Vec b, void *ctx)
 84: {
 85:   UserContext  *user = (UserContext *)ctx;
 86:   PetscInt      i, j, mx, my, xm, ym, xs, ys;
 87:   PetscScalar   Hx, Hy;
 88:   PetscScalar **array;
 89:   DM            da;

 91:   PetscFunctionBeginUser;
 92:   PetscCall(KSPGetDM(ksp, &da));
 93:   PetscCall(DMDAGetInfo(da, 0, &mx, &my, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0));
 94:   Hx = 1.0 / (PetscReal)mx;
 95:   Hy = 1.0 / (PetscReal)my;
 96:   PetscCall(DMDAGetCorners(da, &xs, &ys, 0, &xm, &ym, 0));
 97:   PetscCall(DMDAVecGetArray(da, b, &array));
 98:   for (j = ys; j < ys + ym; j++) {
 99:     for (i = xs; i < xs + xm; i++) array[j][i] = PetscExpScalar(-(((PetscReal)i + 0.5) * Hx) * (((PetscReal)i + 0.5) * Hx) / user->nu) * PetscExpScalar(-(((PetscReal)j + 0.5) * Hy) * (((PetscReal)j + 0.5) * Hy) / user->nu) * Hx * Hy;
100:   }
101:   PetscCall(DMDAVecRestoreArray(da, b, &array));
102:   PetscCall(VecAssemblyBegin(b));
103:   PetscCall(VecAssemblyEnd(b));

105:   /* force right-hand side to be consistent for singular matrix */
106:   /* note this is really a hack, normally the model would provide you with a consistent right handside */
107:   if (user->bcType == NEUMANN) {
108:     MatNullSpace nullspace;

110:     PetscCall(MatNullSpaceCreate(PETSC_COMM_WORLD, PETSC_TRUE, 0, 0, &nullspace));
111:     PetscCall(MatNullSpaceRemove(nullspace, b));
112:     PetscCall(MatNullSpaceDestroy(&nullspace));
113:   }
114:   PetscFunctionReturn(PETSC_SUCCESS);
115: }

117: PetscErrorCode ComputeMatrix(KSP ksp, Mat J, Mat jac, void *ctx)
118: {
119:   UserContext *user = (UserContext *)ctx;
120:   PetscInt     i, j, mx, my, xm, ym, xs, ys, num, numi, numj;
121:   PetscScalar  v[5], Hx, Hy, HydHx, HxdHy;
122:   MatStencil   row, col[5];
123:   DM           da;

125:   PetscFunctionBeginUser;
126:   PetscCall(KSPGetDM(ksp, &da));
127:   PetscCall(DMDAGetInfo(da, 0, &mx, &my, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0));
128:   Hx    = 1.0 / (PetscReal)mx;
129:   Hy    = 1.0 / (PetscReal)my;
130:   HxdHy = Hx / Hy;
131:   HydHx = Hy / Hx;
132:   PetscCall(DMDAGetCorners(da, &xs, &ys, 0, &xm, &ym, 0));
133:   for (j = ys; j < ys + ym; j++) {
134:     for (i = xs; i < xs + xm; i++) {
135:       row.i = i;
136:       row.j = j;
137:       if (i == 0 || j == 0 || i == mx - 1 || j == my - 1) {
138:         if (user->bcType == DIRICHLET) {
139:           v[0] = 2.0 * (HxdHy + HydHx);
140:           PetscCall(MatSetValuesStencil(jac, 1, &row, 1, &row, v, INSERT_VALUES));
141:           SETERRQ(PETSC_COMM_WORLD, PETSC_ERR_SUP, "Dirichlet boundary conditions not supported !");
142:         } else if (user->bcType == NEUMANN) {
143:           num  = 0;
144:           numi = 0;
145:           numj = 0;
146:           if (j != 0) {
147:             v[num]     = -HxdHy;
148:             col[num].i = i;
149:             col[num].j = j - 1;
150:             num++;
151:             numj++;
152:           }
153:           if (i != 0) {
154:             v[num]     = -HydHx;
155:             col[num].i = i - 1;
156:             col[num].j = j;
157:             num++;
158:             numi++;
159:           }
160:           if (i != mx - 1) {
161:             v[num]     = -HydHx;
162:             col[num].i = i + 1;
163:             col[num].j = j;
164:             num++;
165:             numi++;
166:           }
167:           if (j != my - 1) {
168:             v[num]     = -HxdHy;
169:             col[num].i = i;
170:             col[num].j = j + 1;
171:             num++;
172:             numj++;
173:           }
174:           v[num]     = (PetscReal)numj * HxdHy + (PetscReal)numi * HydHx;
175:           col[num].i = i;
176:           col[num].j = j;
177:           num++;
178:           PetscCall(MatSetValuesStencil(jac, 1, &row, num, col, v, INSERT_VALUES));
179:         }
180:       } else {
181:         v[0]     = -HxdHy;
182:         col[0].i = i;
183:         col[0].j = j - 1;
184:         v[1]     = -HydHx;
185:         col[1].i = i - 1;
186:         col[1].j = j;
187:         v[2]     = 2.0 * (HxdHy + HydHx);
188:         col[2].i = i;
189:         col[2].j = j;
190:         v[3]     = -HydHx;
191:         col[3].i = i + 1;
192:         col[3].j = j;
193:         v[4]     = -HxdHy;
194:         col[4].i = i;
195:         col[4].j = j + 1;
196:         PetscCall(MatSetValuesStencil(jac, 1, &row, 5, col, v, INSERT_VALUES));
197:       }
198:     }
199:   }
200:   PetscCall(MatAssemblyBegin(jac, MAT_FINAL_ASSEMBLY));
201:   PetscCall(MatAssemblyEnd(jac, MAT_FINAL_ASSEMBLY));
202:   if (user->bcType == NEUMANN) {
203:     MatNullSpace nullspace;

205:     PetscCall(MatNullSpaceCreate(PETSC_COMM_WORLD, PETSC_TRUE, 0, 0, &nullspace));
206:     PetscCall(MatSetNullSpace(J, nullspace));
207:     PetscCall(MatNullSpaceDestroy(&nullspace));
208:   }
209:   PetscFunctionReturn(PETSC_SUCCESS);
210: }

212: /*TEST

214:    test:
215:      suffix: 2
216:      requires: !single
217:      args: -pc_type mg -pc_mg_levels 5 -ksp_monitor_true_residual -ksp_rtol 1.e-10 -ksp_type cg -mg_levels_pc_type sor -mg_levels_ksp_type richardson -mg_levels_ksp_max_it 2 -mg_coarse_pc_type svd -da_refine 5

219:    test:
220:      suffix: 3
221:      requires: !single
222:      args: -pc_type mg -pc_mg_levels 2 -ksp_monitor_true_residual -ksp_rtol 1.e-10 -ksp_type cg -mg_levels_pc_type sor -mg_levels_ksp_type richardson -mg_levels_ksp_max_it 2 -mg_coarse_pc_type svd -da_refine 5

224:    test:
225:      suffix: 4
226:      requires: !single
227:      args: -pc_type mg -pc_mg_levels 2 -ksp_monitor_true_residual -ksp_rtol 1.e-10 -ksp_type cg -mg_levels_pc_type sor -mg_levels_ksp_type richardson -mg_levels_ksp_max_it 2 -mg_coarse_pc_type svd -da_refine 4

229: TEST*/