static char help[] = "Heat Equation in 2d and 3d with finite elements.\n\ We solve the heat equation in a rectangular\n\ domain, using a parallel unstructured mesh (DMPLEX) to discretize it.\n\ Contributed by: Julian Andrej \n\n\n"; #include #include #include /* Heat equation: du/dt - \Delta u + f = 0 */ typedef enum { SOL_QUADRATIC_LINEAR, SOL_QUADRATIC_TRIG, SOL_TRIG_LINEAR, SOL_TRIG_TRIG, NUM_SOLUTION_TYPES } SolutionType; const char *solutionTypes[NUM_SOLUTION_TYPES + 1] = {"quadratic_linear", "quadratic_trig", "trig_linear", "trig_trig", "unknown"}; typedef struct { SolutionType solType; /* Type of exact solution */ /* Solver setup */ PetscBool expTS; /* Flag for explicit timestepping */ PetscBool lumped; /* Lump the mass matrix */ PetscInt remesh_every; /* Remesh every number of steps */ DM remesh_dm; /* New DM after remeshing */ } AppCtx; /* Exact 2D solution: u = 2t + x^2 + y^2 u_t = 2 \Delta u = 2 + 2 = 4 f = 2 F(u) = 2 - (2 + 2) + 2 = 0 Exact 3D solution: u = 3t + x^2 + y^2 + z^2 F(u) = 3 - (2 + 2 + 2) + 3 = 0 */ static PetscErrorCode mms_quad_lin(PetscInt dim, PetscReal time, const PetscReal x[], PetscInt Nc, PetscScalar *u, PetscCtx ctx) { PetscInt d; *u = dim * time; for (d = 0; d < dim; ++d) *u += x[d] * x[d]; return PETSC_SUCCESS; } static PetscErrorCode mms_quad_lin_t(PetscInt dim, PetscReal time, const PetscReal x[], PetscInt Nc, PetscScalar *u, PetscCtx ctx) { *u = dim; return PETSC_SUCCESS; } static void f0_quad_lin_exp(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar f0[]) { f0[0] = -(PetscScalar)dim; } static void f0_quad_lin(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar f0[]) { PetscScalar exp[1] = {0.}; f0_quad_lin_exp(dim, Nf, NfAux, uOff, uOff_x, u, u_t, u_x, aOff, aOff_x, a, a_t, a_x, t, x, numConstants, constants, exp); f0[0] = u_t[0] - exp[0]; } /* Exact 2D solution: u = 2*cos(t) + x^2 + y^2 F(u) = -2*sint(t) - (2 + 2) + 2*sin(t) + 4 = 0 Exact 3D solution: u = 3*cos(t) + x^2 + y^2 + z^2 F(u) = -3*sin(t) - (2 + 2 + 2) + 3*sin(t) + 6 = 0 */ static PetscErrorCode mms_quad_trig(PetscInt dim, PetscReal time, const PetscReal x[], PetscInt Nc, PetscScalar *u, PetscCtx ctx) { PetscInt d; *u = dim * PetscCosReal(time); for (d = 0; d < dim; ++d) *u += x[d] * x[d]; return PETSC_SUCCESS; } static PetscErrorCode mms_quad_trig_t(PetscInt dim, PetscReal time, const PetscReal x[], PetscInt Nc, PetscScalar *u, PetscCtx ctx) { *u = -dim * PetscSinReal(time); return PETSC_SUCCESS; } static void f0_quad_trig_exp(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar f0[]) { f0[0] = -dim * (PetscSinReal(t) + 2.0); } static void f0_quad_trig(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar f0[]) { PetscScalar exp[1] = {0.}; f0_quad_trig_exp(dim, Nf, NfAux, uOff, uOff_x, u, u_t, u_x, aOff, aOff_x, a, a_t, a_x, t, x, numConstants, constants, exp); f0[0] = u_t[0] - exp[0]; } /* Exact 2D solution: u = 2\pi^2 t + cos(\pi x) + cos(\pi y) F(u) = 2\pi^2 - \pi^2 (cos(\pi x) + cos(\pi y)) + \pi^2 (cos(\pi x) + cos(\pi y)) - 2\pi^2 = 0 Exact 3D solution: u = 3\pi^2 t + cos(\pi x) + cos(\pi y) + cos(\pi z) F(u) = 3\pi^2 - \pi^2 (cos(\pi x) + cos(\pi y) + cos(\pi z)) + \pi^2 (cos(\pi x) + cos(\pi y) + cos(\pi z)) - 3\pi^2 = 0 */ static PetscErrorCode mms_trig_lin(PetscInt dim, PetscReal time, const PetscReal x[], PetscInt Nc, PetscScalar *u, PetscCtx ctx) { PetscInt d; *u = dim * PetscSqr(PETSC_PI) * time; for (d = 0; d < dim; ++d) *u += PetscCosReal(PETSC_PI * x[d]); return PETSC_SUCCESS; } static PetscErrorCode mms_trig_lin_t(PetscInt dim, PetscReal time, const PetscReal x[], PetscInt Nc, PetscScalar *u, PetscCtx ctx) { *u = dim * PetscSqr(PETSC_PI); return PETSC_SUCCESS; } static void f0_trig_lin(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar f0[]) { PetscInt d; f0[0] = u_t[0]; for (d = 0; d < dim; ++d) f0[0] += PetscSqr(PETSC_PI) * (PetscCosReal(PETSC_PI * x[d]) - 1.0); } /* Exact 2D solution: u = pi^2 cos(t) + cos(\pi x) + cos(\pi y) u_t = -pi^2 sin(t) \Delta u = -\pi^2 (cos(\pi x) + cos(\pi y)) f = pi^2 sin(t) - \pi^2 (cos(\pi x) + cos(\pi y)) F(u) = -\pi^2 sin(t) + \pi^2 (cos(\pi x) + cos(\pi y)) - \pi^2 (cos(\pi x) + cos(\pi y)) + \pi^2 sin(t) = 0 Exact 3D solution: u = pi^2 cos(t) + cos(\pi x) + cos(\pi y) + cos(\pi z) u_t = -pi^2 sin(t) \Delta u = -\pi^2 (cos(\pi x) + cos(\pi y) + cos(\pi z)) f = pi^2 sin(t) - \pi^2 (cos(\pi x) + cos(\pi y) + cos(\pi z)) F(u) = -\pi^2 sin(t) + \pi^2 (cos(\pi x) + cos(\pi y) + cos(\pi z)) - \pi^2 (cos(\pi x) + cos(\pi y) + cos(\pi z)) + \pi^2 sin(t) = 0 */ static PetscErrorCode mms_trig_trig(PetscInt dim, PetscReal time, const PetscReal x[], PetscInt Nc, PetscScalar *u, PetscCtx ctx) { PetscInt d; *u = PetscSqr(PETSC_PI) * PetscCosReal(time); for (d = 0; d < dim; ++d) *u += PetscCosReal(PETSC_PI * x[d]); return PETSC_SUCCESS; } static PetscErrorCode mms_trig_trig_t(PetscInt dim, PetscReal time, const PetscReal x[], PetscInt Nc, PetscScalar *u, PetscCtx ctx) { *u = -PetscSqr(PETSC_PI) * PetscSinReal(time); return PETSC_SUCCESS; } static void f0_trig_trig_exp(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar f0[]) { PetscInt d; f0[0] -= PetscSqr(PETSC_PI) * PetscSinReal(t); for (d = 0; d < dim; ++d) f0[0] += PetscSqr(PETSC_PI) * PetscCosReal(PETSC_PI * x[d]); } static void f0_trig_trig(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar f0[]) { PetscScalar exp[1] = {0.}; f0_trig_trig_exp(dim, Nf, NfAux, uOff, uOff_x, u, u_t, u_x, aOff, aOff_x, a, a_t, a_x, t, x, numConstants, constants, exp); f0[0] = u_t[0] - exp[0]; } static void f1_temp_exp(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar f1[]) { for (PetscInt d = 0; d < dim; ++d) f1[d] = -u_x[d]; } static void f1_temp(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar f1[]) { for (PetscInt d = 0; d < dim; ++d) f1[d] = u_x[d]; } static void g3_temp(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, PetscReal u_tShift, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar g3[]) { for (PetscInt d = 0; d < dim; ++d) g3[d * dim + d] = 1.0; } static void g0_temp(PetscInt dim, PetscInt Nf, PetscInt NfAux, const PetscInt uOff[], const PetscInt uOff_x[], const PetscScalar u[], const PetscScalar u_t[], const PetscScalar u_x[], const PetscInt aOff[], const PetscInt aOff_x[], const PetscScalar a[], const PetscScalar a_t[], const PetscScalar a_x[], PetscReal t, PetscReal u_tShift, const PetscReal x[], PetscInt numConstants, const PetscScalar constants[], PetscScalar g0[]) { g0[0] = u_tShift * 1.0; } static PetscErrorCode ProcessOptions(MPI_Comm comm, AppCtx *options) { PetscInt sol; PetscFunctionBeginUser; options->solType = SOL_QUADRATIC_LINEAR; options->expTS = PETSC_FALSE; options->lumped = PETSC_TRUE; options->remesh_every = 0; PetscOptionsBegin(comm, "", "Heat Equation Options", "DMPLEX"); sol = options->solType; PetscCall(PetscOptionsEList("-sol_type", "Type of exact solution", "ex45.c", solutionTypes, NUM_SOLUTION_TYPES, solutionTypes[options->solType], &sol, NULL)); options->solType = (SolutionType)sol; PetscCall(PetscOptionsBool("-explicit", "Use explicit timestepping", "ex45.c", options->expTS, &options->expTS, NULL)); PetscCall(PetscOptionsBool("-lumped", "Lump the mass matrix", "ex45.c", options->lumped, &options->lumped, NULL)); PetscCall(PetscOptionsInt("-remesh_every", "Remesh every number of steps", "ex45.c", options->remesh_every, &options->remesh_every, NULL)); PetscOptionsEnd(); PetscFunctionReturn(PETSC_SUCCESS); } static PetscErrorCode CreateMesh(MPI_Comm comm, DM *dm, AppCtx *ctx) { PetscFunctionBeginUser; PetscCall(DMCreate(comm, dm)); PetscCall(DMSetType(*dm, DMPLEX)); PetscCall(DMSetFromOptions(*dm)); { char convType[256]; PetscBool flg; PetscOptionsBegin(comm, "", "Mesh conversion options", "DMPLEX"); PetscCall(PetscOptionsFList("-dm_plex_convert_type", "Convert DMPlex to another format", __FILE__, DMList, DMPLEX, convType, 256, &flg)); PetscOptionsEnd(); if (flg) { DM dmConv; PetscCall(DMConvert(*dm, convType, &dmConv)); if (dmConv) { PetscCall(DMDestroy(dm)); *dm = dmConv; PetscCall(DMSetFromOptions(*dm)); PetscCall(DMSetUp(*dm)); } } } PetscCall(DMViewFromOptions(*dm, NULL, "-dm_view")); PetscFunctionReturn(PETSC_SUCCESS); } static PetscErrorCode SetupProblem(DM dm, AppCtx *ctx) { PetscDS ds; DMLabel label; const PetscInt id = 1; PetscFunctionBeginUser; PetscCall(DMGetLabel(dm, "marker", &label)); PetscCall(DMGetDS(dm, &ds)); PetscCall(PetscDSSetJacobian(ds, 0, 0, g0_temp, NULL, NULL, g3_temp)); switch (ctx->solType) { case SOL_QUADRATIC_LINEAR: PetscCall(PetscDSSetResidual(ds, 0, f0_quad_lin, f1_temp)); PetscCall(PetscDSSetRHSResidual(ds, 0, f0_quad_lin_exp, f1_temp_exp)); PetscCall(PetscDSSetExactSolution(ds, 0, mms_quad_lin, ctx)); PetscCall(PetscDSSetExactSolutionTimeDerivative(ds, 0, mms_quad_lin_t, ctx)); PetscCall(DMAddBoundary(dm, DM_BC_ESSENTIAL, "wall", label, 1, &id, 0, 0, NULL, (PetscVoidFn *)mms_quad_lin, (PetscVoidFn *)mms_quad_lin_t, ctx, NULL)); break; case SOL_QUADRATIC_TRIG: PetscCall(PetscDSSetResidual(ds, 0, f0_quad_trig, f1_temp)); PetscCall(PetscDSSetRHSResidual(ds, 0, f0_quad_trig_exp, f1_temp_exp)); PetscCall(PetscDSSetExactSolution(ds, 0, mms_quad_trig, ctx)); PetscCall(PetscDSSetExactSolutionTimeDerivative(ds, 0, mms_quad_trig_t, ctx)); PetscCall(DMAddBoundary(dm, DM_BC_ESSENTIAL, "wall", label, 1, &id, 0, 0, NULL, (PetscVoidFn *)mms_quad_trig, (PetscVoidFn *)mms_quad_trig_t, ctx, NULL)); break; case SOL_TRIG_LINEAR: PetscCall(PetscDSSetResidual(ds, 0, f0_trig_lin, f1_temp)); PetscCall(PetscDSSetExactSolution(ds, 0, mms_trig_lin, ctx)); PetscCall(PetscDSSetExactSolutionTimeDerivative(ds, 0, mms_trig_lin_t, ctx)); PetscCall(DMAddBoundary(dm, DM_BC_ESSENTIAL, "wall", label, 1, &id, 0, 0, NULL, (PetscVoidFn *)mms_trig_lin, (PetscVoidFn *)mms_trig_lin_t, ctx, NULL)); break; case SOL_TRIG_TRIG: PetscCall(PetscDSSetResidual(ds, 0, f0_trig_trig, f1_temp)); PetscCall(PetscDSSetRHSResidual(ds, 0, f0_trig_trig_exp, f1_temp_exp)); PetscCall(PetscDSSetExactSolution(ds, 0, mms_trig_trig, ctx)); PetscCall(PetscDSSetExactSolutionTimeDerivative(ds, 0, mms_trig_trig_t, ctx)); break; default: SETERRQ(PetscObjectComm((PetscObject)dm), PETSC_ERR_ARG_WRONG, "Invalid solution type: %s (%d)", solutionTypes[PetscMin(ctx->solType, NUM_SOLUTION_TYPES)], ctx->solType); } PetscFunctionReturn(PETSC_SUCCESS); } static PetscErrorCode SetupDiscretization(DM dm, AppCtx *ctx) { DM plex, cdm = dm; PetscFE fe; PetscBool simplex; PetscInt dim; PetscFunctionBeginUser; PetscCall(DMGetDimension(dm, &dim)); PetscCall(DMConvert(dm, DMPLEX, &plex)); PetscCall(DMPlexIsSimplex(plex, &simplex)); PetscCall(DMDestroy(&plex)); PetscCall(PetscFECreateDefault(PETSC_COMM_SELF, dim, 1, simplex, "temp_", -1, &fe)); PetscCall(PetscObjectSetName((PetscObject)fe, "temperature")); /* Set discretization and boundary conditions for each mesh */ PetscCall(DMSetField(dm, 0, NULL, (PetscObject)fe)); PetscCall(DMCreateDS(dm)); if (ctx->expTS) { PetscDS ds; PetscCall(DMGetDS(dm, &ds)); PetscCall(PetscDSSetImplicit(ds, 0, PETSC_FALSE)); } PetscCall(SetupProblem(dm, ctx)); while (cdm) { PetscCall(DMCopyDisc(dm, cdm)); PetscCall(DMGetCoarseDM(cdm, &cdm)); } PetscCall(PetscFEDestroy(&fe)); PetscFunctionReturn(PETSC_SUCCESS); } #include static PetscErrorCode Remesh(DM dm, Vec U, DM *newdm) { PetscFunctionBeginUser; PetscCall(DMViewFromOptions(dm, NULL, "-remesh_dmin_view")); *newdm = NULL; PetscInt dim; DM plex; PetscBool simplex; PetscCall(DMGetDimension(dm, &dim)); PetscCall(DMConvert(dm, DMPLEX, &plex)); PetscCall(DMPlexIsSimplex(plex, &simplex)); PetscCall(DMDestroy(&plex)); DM dmGrad; PetscFE fe; PetscCall(DMClone(dm, &dmGrad)); PetscCall(PetscFECreateDefault(PETSC_COMM_SELF, dim, dim, simplex, "grad_", -1, &fe)); PetscCall(DMSetField(dmGrad, 0, NULL, (PetscObject)fe)); PetscCall(PetscFEDestroy(&fe)); PetscCall(DMCreateDS(dmGrad)); Vec locU, locG; PetscCall(DMGetLocalVector(dm, &locU)); PetscCall(DMGetLocalVector(dmGrad, &locG)); PetscCall(DMGlobalToLocal(dm, U, INSERT_VALUES, locU)); PetscCall(VecViewFromOptions(locU, NULL, "-remesh_input_view")); PetscCall(DMPlexComputeGradientClementInterpolant(dm, locU, locG)); PetscCall(VecViewFromOptions(locG, NULL, "-remesh_grad_view")); const PetscScalar *g; PetscScalar *u; PetscInt n; PetscCall(VecGetLocalSize(locU, &n)); PetscCall(VecGetArrayWrite(locU, &u)); PetscCall(VecGetArrayRead(locG, &g)); for (PetscInt i = 0; i < n; i++) { PetscReal norm = 0.0; for (PetscInt d = 0; d < dim; d++) norm += PetscSqr(PetscRealPart(g[dim * i + d])); u[i] = PetscSqrtReal(norm); } PetscCall(VecRestoreArrayRead(locG, &g)); PetscCall(VecRestoreArrayWrite(locU, &u)); DM dmM; Vec metric; PetscCall(DMClone(dm, &dmM)); PetscCall(DMPlexMetricCreateIsotropic(dmM, 0, locU, &metric)); PetscCall(DMDestroy(&dmM)); PetscCall(DMRestoreLocalVector(dm, &locU)); PetscCall(DMRestoreLocalVector(dmGrad, &locG)); PetscCall(DMDestroy(&dmGrad)); // TODO remove? PetscScalar scale = 10.0; PetscCall(PetscOptionsGetScalar(NULL, NULL, "-metric_scale", &scale, NULL)); PetscCall(VecScale(metric, scale)); PetscCall(VecViewFromOptions(metric, NULL, "-metric_view")); DMLabel label = NULL; PetscCall(DMGetLabel(dm, "marker", &label)); PetscCall(DMAdaptMetric(dm, metric, label, NULL, newdm)); PetscCall(VecDestroy(&metric)); PetscCall(DMViewFromOptions(*newdm, NULL, "-remesh_dmout_view")); AppCtx *ctx; PetscCall(DMGetApplicationContext(dm, &ctx)); PetscCall(DMSetApplicationContext(*newdm, ctx)); PetscCall(SetupDiscretization(*newdm, ctx)); // TODO ((DM_Plex *)(*newdm)->data)->useHashLocation = ((DM_Plex *)dm->data)->useHashLocation; PetscFunctionReturn(PETSC_SUCCESS); } static PetscErrorCode SetInitialConditions(TS ts, Vec u) { DM dm; PetscReal t; PetscFunctionBeginUser; PetscCall(TSGetDM(ts, &dm)); PetscCall(TSGetTime(ts, &t)); PetscCall(DMComputeExactSolution(dm, t, u, NULL)); PetscCall(VecSetOptionsPrefix(u, NULL)); PetscFunctionReturn(PETSC_SUCCESS); } static PetscErrorCode TransferSetUp(TS ts, PetscInt step, PetscReal time, Vec U, PetscBool *remesh, void *tctx) { AppCtx *ctx = (AppCtx *)tctx; PetscFunctionBeginUser; *remesh = PETSC_FALSE; if (ctx->remesh_every > 0 && step && step % ctx->remesh_every == 0) { DM dm; *remesh = PETSC_TRUE; PetscCall(TSGetDM(ts, &dm)); PetscCall(Remesh(dm, U, &ctx->remesh_dm)); } PetscFunctionReturn(PETSC_SUCCESS); } static PetscErrorCode TransferVecs(TS ts, PetscInt nv, Vec vin[], Vec vout[], void *tctx) { AppCtx *ctx = (AppCtx *)tctx; DM dm; Mat Interp; PetscFunctionBeginUser; PetscCall(TSGetDM(ts, &dm)); PetscCall(DMCreateInterpolation(dm, ctx->remesh_dm, &Interp, NULL)); for (PetscInt i = 0; i < nv; i++) { PetscCall(DMCreateGlobalVector(ctx->remesh_dm, &vout[i])); PetscCall(MatMult(Interp, vin[i], vout[i])); } PetscCall(MatDestroy(&Interp)); PetscCall(TSSetDM(ts, ctx->remesh_dm)); PetscCall(DMDestroy(&ctx->remesh_dm)); PetscFunctionReturn(PETSC_SUCCESS); } int main(int argc, char **argv) { DM dm; TS ts; Vec u; AppCtx ctx; PetscFunctionBeginUser; PetscCall(PetscInitialize(&argc, &argv, NULL, help)); PetscCall(ProcessOptions(PETSC_COMM_WORLD, &ctx)); PetscCall(CreateMesh(PETSC_COMM_WORLD, &dm, &ctx)); PetscCall(DMSetApplicationContext(dm, &ctx)); PetscCall(SetupDiscretization(dm, &ctx)); PetscCall(TSCreate(PETSC_COMM_WORLD, &ts)); PetscCall(TSSetDM(ts, dm)); PetscCall(DMTSSetBoundaryLocal(dm, DMPlexTSComputeBoundary, &ctx)); if (ctx.expTS) { PetscCall(DMTSSetRHSFunctionLocal(dm, DMPlexTSComputeRHSFunctionFEM, &ctx)); if (ctx.lumped) PetscCall(DMTSCreateRHSMassMatrixLumped(dm)); else PetscCall(DMTSCreateRHSMassMatrix(dm)); } else { PetscCall(DMTSSetIFunctionLocal(dm, DMPlexTSComputeIFunctionFEM, &ctx)); PetscCall(DMTSSetIJacobianLocal(dm, DMPlexTSComputeIJacobianFEM, &ctx)); } PetscCall(TSSetMaxTime(ts, 1.0)); PetscCall(TSSetExactFinalTime(ts, TS_EXACTFINALTIME_MATCHSTEP)); PetscCall(TSSetFromOptions(ts)); PetscCall(TSSetComputeInitialCondition(ts, SetInitialConditions)); PetscCall(TSSetResize(ts, PETSC_FALSE, TransferSetUp, TransferVecs, &ctx)); PetscCall(DMCreateGlobalVector(dm, &u)); PetscCall(DMTSCheckFromOptions(ts, u)); PetscCall(SetInitialConditions(ts, u)); PetscCall(PetscObjectSetName((PetscObject)u, "temperature")); PetscCall(TSSetSolution(ts, u)); PetscCall(VecViewFromOptions(u, NULL, "-u0_view")); PetscCall(VecDestroy(&u)); PetscCall(TSSolve(ts, NULL)); PetscCall(TSGetSolution(ts, &u)); PetscCall(VecViewFromOptions(u, NULL, "-uf_view")); PetscCall(DMTSCheckFromOptions(ts, u)); if (ctx.expTS) PetscCall(DMTSDestroyRHSMassMatrix(dm)); PetscCall(TSDestroy(&ts)); PetscCall(DMDestroy(&dm)); PetscCall(PetscFinalize()); return 0; } /*TEST test: suffix: 2d_p1 requires: triangle args: -sol_type quadratic_linear -dm_refine 1 -temp_petscspace_degree 1 -dmts_check .0001 \ -ts_type beuler -ts_max_steps 5 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: suffix: 2d_p1_exp requires: triangle args: -sol_type quadratic_linear -dm_refine 1 -temp_petscspace_degree 1 -explicit \ -ts_type euler -ts_max_steps 4 -ts_time_step 1e-3 -ts_monitor_error test: # -dm_refine 2 -convest_num_refine 3 get L_2 convergence rate: [1.9] suffix: 2d_p1_sconv requires: triangle args: -sol_type quadratic_linear -temp_petscspace_degree 1 -ts_convergence_estimate -ts_convergence_temporal 0 -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 1 -ts_time_step 0.00001 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 2 -convest_num_refine 3 get L_2 convergence rate: [2.1] suffix: 2d_p1_sconv_2 requires: triangle args: -sol_type trig_trig -temp_petscspace_degree 1 -ts_convergence_estimate -ts_convergence_temporal 0 -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 1 -ts_time_step 1e-6 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 4 -convest_num_refine 3 get L_2 convergence rate: [1.2] suffix: 2d_p1_tconv requires: triangle args: -sol_type quadratic_trig -temp_petscspace_degree 1 -ts_convergence_estimate -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 4 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 6 -convest_num_refine 3 get L_2 convergence rate: [1.0] suffix: 2d_p1_tconv_2 requires: triangle args: -sol_type trig_trig -temp_petscspace_degree 1 -ts_convergence_estimate -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 4 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: # The L_2 convergence rate cannot be seen since stability of the explicit integrator requires that is be more accurate than the grid suffix: 2d_p1_exp_tconv_2 requires: triangle args: -sol_type trig_trig -temp_petscspace_degree 1 -explicit -ts_convergence_estimate -convest_num_refine 1 \ -ts_type euler -ts_max_steps 4 -ts_time_step 1e-4 -lumped 0 -mass_pc_type lu test: # The L_2 convergence rate cannot be seen since stability of the explicit integrator requires that is be more accurate than the grid suffix: 2d_p1_exp_tconv_2_lump requires: triangle args: -sol_type trig_trig -temp_petscspace_degree 1 -explicit -ts_convergence_estimate -convest_num_refine 1 \ -ts_type euler -ts_max_steps 4 -ts_time_step 1e-4 test: suffix: 2d_p2 requires: triangle args: -sol_type quadratic_linear -dm_refine 0 -temp_petscspace_degree 2 -dmts_check .0001 \ -ts_type beuler -ts_max_steps 5 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 2 -convest_num_refine 3 get L_2 convergence rate: [2.9] suffix: 2d_p2_sconv requires: triangle args: -sol_type trig_linear -temp_petscspace_degree 2 -ts_convergence_estimate -ts_convergence_temporal 0 -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 1 -ts_time_step 0.00000001 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 2 -convest_num_refine 3 get L_2 convergence rate: [3.1] suffix: 2d_p2_sconv_2 requires: triangle args: -sol_type trig_trig -temp_petscspace_degree 2 -ts_convergence_estimate -ts_convergence_temporal 0 -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 1 -ts_time_step 0.00000001 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 3 -convest_num_refine 3 get L_2 convergence rate: [1.0] suffix: 2d_p2_tconv requires: triangle args: -sol_type quadratic_trig -temp_petscspace_degree 2 -ts_convergence_estimate -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 4 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 3 -convest_num_refine 3 get L_2 convergence rate: [1.0] suffix: 2d_p2_tconv_2 requires: triangle args: -sol_type trig_trig -temp_petscspace_degree 2 -ts_convergence_estimate -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 4 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: suffix: 2d_q1 args: -sol_type quadratic_linear -dm_plex_simplex 0 -dm_refine 1 -temp_petscspace_degree 1 -dmts_check .0001 \ -ts_type beuler -ts_max_steps 5 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 2 -convest_num_refine 3 get L_2 convergence rate: [1.9] suffix: 2d_q1_sconv args: -sol_type quadratic_linear -dm_plex_simplex 0 -temp_petscspace_degree 1 -ts_convergence_estimate -ts_convergence_temporal 0 -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 1 -ts_time_step 0.00001 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 4 -convest_num_refine 3 get L_2 convergence rate: [1.2] suffix: 2d_q1_tconv args: -sol_type quadratic_trig -dm_plex_simplex 0 -temp_petscspace_degree 1 -ts_convergence_estimate -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 4 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: suffix: 2d_q2 args: -sol_type quadratic_linear -dm_plex_simplex 0 -dm_refine 0 -temp_petscspace_degree 2 -dmts_check .0001 \ -ts_type beuler -ts_max_steps 5 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 2 -convest_num_refine 3 get L_2 convergence rate: [2.9] suffix: 2d_q2_sconv args: -sol_type trig_linear -dm_plex_simplex 0 -temp_petscspace_degree 2 -ts_convergence_estimate -ts_convergence_temporal 0 -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 1 -ts_time_step 0.00000001 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 3 -convest_num_refine 3 get L_2 convergence rate: [1.0] suffix: 2d_q2_tconv args: -sol_type quadratic_trig -dm_plex_simplex 0 -temp_petscspace_degree 2 -ts_convergence_estimate -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 4 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: suffix: 3d_p1 requires: ctetgen args: -sol_type quadratic_linear -dm_refine 1 -temp_petscspace_degree 1 -dmts_check .0001 \ -ts_type beuler -ts_max_steps 5 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 2 -convest_num_refine 3 get L_2 convergence rate: [1.9] suffix: 3d_p1_sconv requires: ctetgen args: -sol_type quadratic_linear -temp_petscspace_degree 1 -ts_convergence_estimate -ts_convergence_temporal 0 -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 1 -ts_time_step 0.00001 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 4 -convest_num_refine 3 get L_2 convergence rate: [1.2] suffix: 3d_p1_tconv requires: ctetgen args: -sol_type quadratic_trig -temp_petscspace_degree 1 -ts_convergence_estimate -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 4 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: suffix: 3d_p2 requires: ctetgen args: -sol_type quadratic_linear -dm_refine 0 -temp_petscspace_degree 2 -dmts_check .0001 \ -ts_type beuler -ts_max_steps 5 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 2 -convest_num_refine 3 get L_2 convergence rate: [2.9] suffix: 3d_p2_sconv requires: ctetgen args: -sol_type trig_linear -temp_petscspace_degree 2 -ts_convergence_estimate -ts_convergence_temporal 0 -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 1 -ts_time_step 0.00000001 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 3 -convest_num_refine 3 get L_2 convergence rate: [1.0] suffix: 3d_p2_tconv requires: ctetgen args: -sol_type quadratic_trig -temp_petscspace_degree 2 -ts_convergence_estimate -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 4 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: suffix: 3d_q1 args: -sol_type quadratic_linear -dm_plex_simplex 0 -dm_refine 1 -temp_petscspace_degree 1 -dmts_check .0001 \ -ts_type beuler -ts_max_steps 5 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 2 -convest_num_refine 3 get L_2 convergence rate: [1.9] suffix: 3d_q1_sconv args: -sol_type quadratic_linear -dm_plex_simplex 0 -temp_petscspace_degree 1 -ts_convergence_estimate -ts_convergence_temporal 0 -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 1 -ts_time_step 0.00001 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 4 -convest_num_refine 3 get L_2 convergence rate: [1.2] suffix: 3d_q1_tconv args: -sol_type quadratic_trig -dm_plex_simplex 0 -temp_petscspace_degree 1 -ts_convergence_estimate -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 4 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: suffix: 3d_q2 args: -sol_type quadratic_linear -dm_plex_simplex 0 -dm_refine 0 -temp_petscspace_degree 2 -dmts_check .0001 \ -ts_type beuler -ts_max_steps 5 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 2 -convest_num_refine 3 get L_2 convergence rate: [2.9] suffix: 3d_q2_sconv args: -sol_type trig_linear -dm_plex_simplex 0 -temp_petscspace_degree 2 -ts_convergence_estimate -ts_convergence_temporal 0 -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 1 -ts_time_step 0.00000001 -snes_error_if_not_converged -pc_type lu test: # -dm_refine 3 -convest_num_refine 3 get L_2 convergence rate: [1.0] suffix: 3d_q2_tconv args: -sol_type quadratic_trig -dm_plex_simplex 0 -temp_petscspace_degree 2 -ts_convergence_estimate -convest_num_refine 1 \ -ts_type beuler -ts_max_steps 4 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: # For a nice picture, -bd_dm_refine 2 -dm_refine 1 -dm_view hdf5:${PETSC_DIR}/sol.h5 -ts_monitor_solution hdf5:${PETSC_DIR}/sol.h5::append suffix: egads_sphere requires: egads ctetgen datafilespath args: -sol_type quadratic_linear \ -dm_plex_boundary_filename ${DATAFILESPATH}/meshes/cad/sphere_example.egadslite -dm_plex_boundary_label marker \ -temp_petscspace_degree 2 -dmts_check .0001 \ -ts_type beuler -ts_max_steps 5 -ts_time_step 0.1 -snes_error_if_not_converged -pc_type lu test: suffix: remesh requires: triangle mmg args: -sol_type quadratic_trig -dm_refine 2 -temp_petscspace_degree 1 -ts_type beuler -ts_time_step 0.01 -snes_error_if_not_converged -pc_type lu -grad_petscspace_degree 1 -dm_adaptor mmg -dm_plex_hash_location -remesh_every 5 output_file: output/empty.out TEST*/