1*c4762a1bSJed Brown 2*c4762a1bSJed Brown static char help[] ="Solves a simple time-dependent linear PDE (the heat equation).\n\ 3*c4762a1bSJed Brown Input parameters include:\n\ 4*c4762a1bSJed Brown -m <points>, where <points> = number of grid points\n\ 5*c4762a1bSJed Brown -time_dependent_rhs : Treat the problem as having a time-dependent right-hand side\n\ 6*c4762a1bSJed Brown -debug : Activate debugging printouts\n\ 7*c4762a1bSJed Brown -nox : Deactivate x-window graphics\n\n"; 8*c4762a1bSJed Brown 9*c4762a1bSJed Brown /* 10*c4762a1bSJed Brown Concepts: TS^time-dependent linear problems 11*c4762a1bSJed Brown Concepts: TS^heat equation 12*c4762a1bSJed Brown Concepts: TS^diffusion equation 13*c4762a1bSJed Brown Processors: n 14*c4762a1bSJed Brown */ 15*c4762a1bSJed Brown 16*c4762a1bSJed Brown /* ------------------------------------------------------------------------ 17*c4762a1bSJed Brown 18*c4762a1bSJed Brown This program solves the one-dimensional heat equation (also called the 19*c4762a1bSJed Brown diffusion equation), 20*c4762a1bSJed Brown u_t = u_xx, 21*c4762a1bSJed Brown on the domain 0 <= x <= 1, with the boundary conditions 22*c4762a1bSJed Brown u(t,0) = 0, u(t,1) = 0, 23*c4762a1bSJed Brown and the initial condition 24*c4762a1bSJed Brown u(0,x) = sin(6*pi*x) + 3*sin(2*pi*x). 25*c4762a1bSJed Brown This is a linear, second-order, parabolic equation. 26*c4762a1bSJed Brown 27*c4762a1bSJed Brown We discretize the right-hand side using finite differences with 28*c4762a1bSJed Brown uniform grid spacing h: 29*c4762a1bSJed Brown u_xx = (u_{i+1} - 2u_{i} + u_{i-1})/(h^2) 30*c4762a1bSJed Brown We then demonstrate time evolution using the various TS methods by 31*c4762a1bSJed Brown running the program via 32*c4762a1bSJed Brown mpiexec -n <procs> ex3 -ts_type <timestepping solver> 33*c4762a1bSJed Brown 34*c4762a1bSJed Brown We compare the approximate solution with the exact solution, given by 35*c4762a1bSJed Brown u_exact(x,t) = exp(-36*pi*pi*t) * sin(6*pi*x) + 36*c4762a1bSJed Brown 3*exp(-4*pi*pi*t) * sin(2*pi*x) 37*c4762a1bSJed Brown 38*c4762a1bSJed Brown Notes: 39*c4762a1bSJed Brown This code demonstrates the TS solver interface to two variants of 40*c4762a1bSJed Brown linear problems, u_t = f(u,t), namely 41*c4762a1bSJed Brown - time-dependent f: f(u,t) is a function of t 42*c4762a1bSJed Brown - time-independent f: f(u,t) is simply f(u) 43*c4762a1bSJed Brown 44*c4762a1bSJed Brown The uniprocessor version of this code is ts/tutorials/ex3.c 45*c4762a1bSJed Brown 46*c4762a1bSJed Brown ------------------------------------------------------------------------- */ 47*c4762a1bSJed Brown 48*c4762a1bSJed Brown /* 49*c4762a1bSJed Brown Include "petscdmda.h" so that we can use distributed arrays (DMDAs) to manage 50*c4762a1bSJed Brown the parallel grid. Include "petscts.h" so that we can use TS solvers. 51*c4762a1bSJed Brown Note that this file automatically includes: 52*c4762a1bSJed Brown petscsys.h - base PETSc routines petscvec.h - vectors 53*c4762a1bSJed Brown petscmat.h - matrices 54*c4762a1bSJed Brown petscis.h - index sets petscksp.h - Krylov subspace methods 55*c4762a1bSJed Brown petscviewer.h - viewers petscpc.h - preconditioners 56*c4762a1bSJed Brown petscksp.h - linear solvers petscsnes.h - nonlinear solvers 57*c4762a1bSJed Brown */ 58*c4762a1bSJed Brown 59*c4762a1bSJed Brown #include <petscdm.h> 60*c4762a1bSJed Brown #include <petscdmda.h> 61*c4762a1bSJed Brown #include <petscts.h> 62*c4762a1bSJed Brown #include <petscdraw.h> 63*c4762a1bSJed Brown 64*c4762a1bSJed Brown /* 65*c4762a1bSJed Brown User-defined application context - contains data needed by the 66*c4762a1bSJed Brown application-provided call-back routines. 67*c4762a1bSJed Brown */ 68*c4762a1bSJed Brown typedef struct { 69*c4762a1bSJed Brown MPI_Comm comm; /* communicator */ 70*c4762a1bSJed Brown DM da; /* distributed array data structure */ 71*c4762a1bSJed Brown Vec localwork; /* local ghosted work vector */ 72*c4762a1bSJed Brown Vec u_local; /* local ghosted approximate solution vector */ 73*c4762a1bSJed Brown Vec solution; /* global exact solution vector */ 74*c4762a1bSJed Brown PetscInt m; /* total number of grid points */ 75*c4762a1bSJed Brown PetscReal h; /* mesh width h = 1/(m-1) */ 76*c4762a1bSJed Brown PetscBool debug; /* flag (1 indicates activation of debugging printouts) */ 77*c4762a1bSJed Brown PetscViewer viewer1,viewer2; /* viewers for the solution and error */ 78*c4762a1bSJed Brown PetscReal norm_2,norm_max; /* error norms */ 79*c4762a1bSJed Brown } AppCtx; 80*c4762a1bSJed Brown 81*c4762a1bSJed Brown /* 82*c4762a1bSJed Brown User-defined routines 83*c4762a1bSJed Brown */ 84*c4762a1bSJed Brown extern PetscErrorCode InitialConditions(Vec,AppCtx*); 85*c4762a1bSJed Brown extern PetscErrorCode RHSMatrixHeat(TS,PetscReal,Vec,Mat,Mat,void*); 86*c4762a1bSJed Brown extern PetscErrorCode RHSFunctionHeat(TS,PetscReal,Vec,Vec,void*); 87*c4762a1bSJed Brown extern PetscErrorCode Monitor(TS,PetscInt,PetscReal,Vec,void*); 88*c4762a1bSJed Brown extern PetscErrorCode ExactSolution(PetscReal,Vec,AppCtx*); 89*c4762a1bSJed Brown 90*c4762a1bSJed Brown int main(int argc,char **argv) 91*c4762a1bSJed Brown { 92*c4762a1bSJed Brown AppCtx appctx; /* user-defined application context */ 93*c4762a1bSJed Brown TS ts; /* timestepping context */ 94*c4762a1bSJed Brown Mat A; /* matrix data structure */ 95*c4762a1bSJed Brown Vec u; /* approximate solution vector */ 96*c4762a1bSJed Brown PetscReal time_total_max = 1.0; /* default max total time */ 97*c4762a1bSJed Brown PetscInt time_steps_max = 100; /* default max timesteps */ 98*c4762a1bSJed Brown PetscDraw draw; /* drawing context */ 99*c4762a1bSJed Brown PetscErrorCode ierr; 100*c4762a1bSJed Brown PetscInt steps,m; 101*c4762a1bSJed Brown PetscMPIInt size; 102*c4762a1bSJed Brown PetscReal dt,ftime; 103*c4762a1bSJed Brown PetscBool flg; 104*c4762a1bSJed Brown TSProblemType tsproblem = TS_LINEAR; 105*c4762a1bSJed Brown 106*c4762a1bSJed Brown /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 107*c4762a1bSJed Brown Initialize program and set problem parameters 108*c4762a1bSJed Brown - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */ 109*c4762a1bSJed Brown 110*c4762a1bSJed Brown ierr = PetscInitialize(&argc,&argv,(char*)0,help);if (ierr) return ierr; 111*c4762a1bSJed Brown appctx.comm = PETSC_COMM_WORLD; 112*c4762a1bSJed Brown 113*c4762a1bSJed Brown m = 60; 114*c4762a1bSJed Brown ierr = PetscOptionsGetInt(NULL,NULL,"-m",&m,NULL);CHKERRQ(ierr); 115*c4762a1bSJed Brown ierr = PetscOptionsHasName(NULL,NULL,"-debug",&appctx.debug);CHKERRQ(ierr); 116*c4762a1bSJed Brown appctx.m = m; 117*c4762a1bSJed Brown appctx.h = 1.0/(m-1.0); 118*c4762a1bSJed Brown appctx.norm_2 = 0.0; 119*c4762a1bSJed Brown appctx.norm_max = 0.0; 120*c4762a1bSJed Brown 121*c4762a1bSJed Brown ierr = MPI_Comm_size(PETSC_COMM_WORLD,&size);CHKERRQ(ierr); 122*c4762a1bSJed Brown ierr = PetscPrintf(PETSC_COMM_WORLD,"Solving a linear TS problem, number of processors = %d\n",size);CHKERRQ(ierr); 123*c4762a1bSJed Brown 124*c4762a1bSJed Brown /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 125*c4762a1bSJed Brown Create vector data structures 126*c4762a1bSJed Brown - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */ 127*c4762a1bSJed Brown /* 128*c4762a1bSJed Brown Create distributed array (DMDA) to manage parallel grid and vectors 129*c4762a1bSJed Brown and to set up the ghost point communication pattern. There are M 130*c4762a1bSJed Brown total grid values spread equally among all the processors. 131*c4762a1bSJed Brown */ 132*c4762a1bSJed Brown 133*c4762a1bSJed Brown ierr = DMDACreate1d(PETSC_COMM_WORLD,DM_BOUNDARY_NONE,m,1,1,NULL,&appctx.da);CHKERRQ(ierr); 134*c4762a1bSJed Brown ierr = DMSetFromOptions(appctx.da);CHKERRQ(ierr); 135*c4762a1bSJed Brown ierr = DMSetUp(appctx.da);CHKERRQ(ierr); 136*c4762a1bSJed Brown 137*c4762a1bSJed Brown /* 138*c4762a1bSJed Brown Extract global and local vectors from DMDA; we use these to store the 139*c4762a1bSJed Brown approximate solution. Then duplicate these for remaining vectors that 140*c4762a1bSJed Brown have the same types. 141*c4762a1bSJed Brown */ 142*c4762a1bSJed Brown ierr = DMCreateGlobalVector(appctx.da,&u);CHKERRQ(ierr); 143*c4762a1bSJed Brown ierr = DMCreateLocalVector(appctx.da,&appctx.u_local);CHKERRQ(ierr); 144*c4762a1bSJed Brown 145*c4762a1bSJed Brown /* 146*c4762a1bSJed Brown Create local work vector for use in evaluating right-hand-side function; 147*c4762a1bSJed Brown create global work vector for storing exact solution. 148*c4762a1bSJed Brown */ 149*c4762a1bSJed Brown ierr = VecDuplicate(appctx.u_local,&appctx.localwork);CHKERRQ(ierr); 150*c4762a1bSJed Brown ierr = VecDuplicate(u,&appctx.solution);CHKERRQ(ierr); 151*c4762a1bSJed Brown 152*c4762a1bSJed Brown /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 153*c4762a1bSJed Brown Set up displays to show graphs of the solution and error 154*c4762a1bSJed Brown - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */ 155*c4762a1bSJed Brown 156*c4762a1bSJed Brown ierr = PetscViewerDrawOpen(PETSC_COMM_WORLD,0,"",80,380,400,160,&appctx.viewer1);CHKERRQ(ierr); 157*c4762a1bSJed Brown ierr = PetscViewerDrawGetDraw(appctx.viewer1,0,&draw);CHKERRQ(ierr); 158*c4762a1bSJed Brown ierr = PetscDrawSetDoubleBuffer(draw);CHKERRQ(ierr); 159*c4762a1bSJed Brown ierr = PetscViewerDrawOpen(PETSC_COMM_WORLD,0,"",80,0,400,160,&appctx.viewer2);CHKERRQ(ierr); 160*c4762a1bSJed Brown ierr = PetscViewerDrawGetDraw(appctx.viewer2,0,&draw);CHKERRQ(ierr); 161*c4762a1bSJed Brown ierr = PetscDrawSetDoubleBuffer(draw);CHKERRQ(ierr); 162*c4762a1bSJed Brown 163*c4762a1bSJed Brown /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 164*c4762a1bSJed Brown Create timestepping solver context 165*c4762a1bSJed Brown - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */ 166*c4762a1bSJed Brown 167*c4762a1bSJed Brown ierr = TSCreate(PETSC_COMM_WORLD,&ts);CHKERRQ(ierr); 168*c4762a1bSJed Brown 169*c4762a1bSJed Brown flg = PETSC_FALSE; 170*c4762a1bSJed Brown ierr = PetscOptionsGetBool(NULL,NULL,"-nonlinear",&flg,NULL);CHKERRQ(ierr); 171*c4762a1bSJed Brown ierr = TSSetProblemType(ts,flg ? TS_NONLINEAR : TS_LINEAR);CHKERRQ(ierr); 172*c4762a1bSJed Brown 173*c4762a1bSJed Brown /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 174*c4762a1bSJed Brown Set optional user-defined monitoring routine 175*c4762a1bSJed Brown - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */ 176*c4762a1bSJed Brown ierr = TSMonitorSet(ts,Monitor,&appctx,NULL);CHKERRQ(ierr); 177*c4762a1bSJed Brown 178*c4762a1bSJed Brown /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 179*c4762a1bSJed Brown 180*c4762a1bSJed Brown Create matrix data structure; set matrix evaluation routine. 181*c4762a1bSJed Brown - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */ 182*c4762a1bSJed Brown 183*c4762a1bSJed Brown ierr = MatCreate(PETSC_COMM_WORLD,&A);CHKERRQ(ierr); 184*c4762a1bSJed Brown ierr = MatSetSizes(A,PETSC_DECIDE,PETSC_DECIDE,m,m);CHKERRQ(ierr); 185*c4762a1bSJed Brown ierr = MatSetFromOptions(A);CHKERRQ(ierr); 186*c4762a1bSJed Brown ierr = MatSetUp(A);CHKERRQ(ierr); 187*c4762a1bSJed Brown 188*c4762a1bSJed Brown flg = PETSC_FALSE; 189*c4762a1bSJed Brown ierr = PetscOptionsGetBool(NULL,NULL,"-time_dependent_rhs",&flg,NULL);CHKERRQ(ierr); 190*c4762a1bSJed Brown if (flg) { 191*c4762a1bSJed Brown /* 192*c4762a1bSJed Brown For linear problems with a time-dependent f(u,t) in the equation 193*c4762a1bSJed Brown u_t = f(u,t), the user provides the discretized right-hand-side 194*c4762a1bSJed Brown as a time-dependent matrix. 195*c4762a1bSJed Brown */ 196*c4762a1bSJed Brown ierr = TSSetRHSFunction(ts,NULL,TSComputeRHSFunctionLinear,&appctx);CHKERRQ(ierr); 197*c4762a1bSJed Brown ierr = TSSetRHSJacobian(ts,A,A,RHSMatrixHeat,&appctx);CHKERRQ(ierr); 198*c4762a1bSJed Brown } else { 199*c4762a1bSJed Brown /* 200*c4762a1bSJed Brown For linear problems with a time-independent f(u) in the equation 201*c4762a1bSJed Brown u_t = f(u), the user provides the discretized right-hand-side 202*c4762a1bSJed Brown as a matrix only once, and then sets a null matrix evaluation 203*c4762a1bSJed Brown routine. 204*c4762a1bSJed Brown */ 205*c4762a1bSJed Brown ierr = RHSMatrixHeat(ts,0.0,u,A,A,&appctx);CHKERRQ(ierr); 206*c4762a1bSJed Brown ierr = TSSetRHSFunction(ts,NULL,TSComputeRHSFunctionLinear,&appctx);CHKERRQ(ierr); 207*c4762a1bSJed Brown ierr = TSSetRHSJacobian(ts,A,A,TSComputeRHSJacobianConstant,&appctx);CHKERRQ(ierr); 208*c4762a1bSJed Brown } 209*c4762a1bSJed Brown 210*c4762a1bSJed Brown if (tsproblem == TS_NONLINEAR) { 211*c4762a1bSJed Brown SNES snes; 212*c4762a1bSJed Brown ierr = TSSetRHSFunction(ts,NULL,RHSFunctionHeat,&appctx);CHKERRQ(ierr); 213*c4762a1bSJed Brown ierr = TSGetSNES(ts,&snes);CHKERRQ(ierr); 214*c4762a1bSJed Brown ierr = SNESSetJacobian(snes,NULL,NULL,SNESComputeJacobianDefault,NULL);CHKERRQ(ierr); 215*c4762a1bSJed Brown } 216*c4762a1bSJed Brown 217*c4762a1bSJed Brown /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 218*c4762a1bSJed Brown Set solution vector and initial timestep 219*c4762a1bSJed Brown - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */ 220*c4762a1bSJed Brown 221*c4762a1bSJed Brown dt = appctx.h*appctx.h/2.0; 222*c4762a1bSJed Brown ierr = TSSetTimeStep(ts,dt);CHKERRQ(ierr); 223*c4762a1bSJed Brown ierr = TSSetSolution(ts,u);CHKERRQ(ierr); 224*c4762a1bSJed Brown 225*c4762a1bSJed Brown /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 226*c4762a1bSJed Brown Customize timestepping solver: 227*c4762a1bSJed Brown - Set the solution method to be the Backward Euler method. 228*c4762a1bSJed Brown - Set timestepping duration info 229*c4762a1bSJed Brown Then set runtime options, which can override these defaults. 230*c4762a1bSJed Brown For example, 231*c4762a1bSJed Brown -ts_max_steps <maxsteps> -ts_max_time <maxtime> 232*c4762a1bSJed Brown to override the defaults set by TSSetMaxSteps()/TSSetMaxTime(). 233*c4762a1bSJed Brown - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */ 234*c4762a1bSJed Brown 235*c4762a1bSJed Brown ierr = TSSetMaxSteps(ts,time_steps_max);CHKERRQ(ierr); 236*c4762a1bSJed Brown ierr = TSSetMaxTime(ts,time_total_max);CHKERRQ(ierr); 237*c4762a1bSJed Brown ierr = TSSetExactFinalTime(ts,TS_EXACTFINALTIME_STEPOVER);CHKERRQ(ierr); 238*c4762a1bSJed Brown ierr = TSSetFromOptions(ts);CHKERRQ(ierr); 239*c4762a1bSJed Brown 240*c4762a1bSJed Brown /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 241*c4762a1bSJed Brown Solve the problem 242*c4762a1bSJed Brown - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */ 243*c4762a1bSJed Brown 244*c4762a1bSJed Brown /* 245*c4762a1bSJed Brown Evaluate initial conditions 246*c4762a1bSJed Brown */ 247*c4762a1bSJed Brown ierr = InitialConditions(u,&appctx);CHKERRQ(ierr); 248*c4762a1bSJed Brown 249*c4762a1bSJed Brown /* 250*c4762a1bSJed Brown Run the timestepping solver 251*c4762a1bSJed Brown */ 252*c4762a1bSJed Brown ierr = TSSolve(ts,u);CHKERRQ(ierr); 253*c4762a1bSJed Brown ierr = TSGetSolveTime(ts,&ftime);CHKERRQ(ierr); 254*c4762a1bSJed Brown ierr = TSGetStepNumber(ts,&steps);CHKERRQ(ierr); 255*c4762a1bSJed Brown 256*c4762a1bSJed Brown /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 257*c4762a1bSJed Brown View timestepping solver info 258*c4762a1bSJed Brown - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */ 259*c4762a1bSJed Brown ierr = PetscPrintf(PETSC_COMM_WORLD,"Total timesteps %D, Final time %g\n",steps,(double)ftime);CHKERRQ(ierr); 260*c4762a1bSJed Brown ierr = PetscPrintf(PETSC_COMM_WORLD,"Avg. error (2 norm) = %g Avg. error (max norm) = %g\n",(double)(appctx.norm_2/steps),(double)(appctx.norm_max/steps));CHKERRQ(ierr); 261*c4762a1bSJed Brown 262*c4762a1bSJed Brown /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 263*c4762a1bSJed Brown Free work space. All PETSc objects should be destroyed when they 264*c4762a1bSJed Brown are no longer needed. 265*c4762a1bSJed Brown - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */ 266*c4762a1bSJed Brown 267*c4762a1bSJed Brown ierr = TSDestroy(&ts);CHKERRQ(ierr); 268*c4762a1bSJed Brown ierr = MatDestroy(&A);CHKERRQ(ierr); 269*c4762a1bSJed Brown ierr = VecDestroy(&u);CHKERRQ(ierr); 270*c4762a1bSJed Brown ierr = PetscViewerDestroy(&appctx.viewer1);CHKERRQ(ierr); 271*c4762a1bSJed Brown ierr = PetscViewerDestroy(&appctx.viewer2);CHKERRQ(ierr); 272*c4762a1bSJed Brown ierr = VecDestroy(&appctx.localwork);CHKERRQ(ierr); 273*c4762a1bSJed Brown ierr = VecDestroy(&appctx.solution);CHKERRQ(ierr); 274*c4762a1bSJed Brown ierr = VecDestroy(&appctx.u_local);CHKERRQ(ierr); 275*c4762a1bSJed Brown ierr = DMDestroy(&appctx.da);CHKERRQ(ierr); 276*c4762a1bSJed Brown 277*c4762a1bSJed Brown /* 278*c4762a1bSJed Brown Always call PetscFinalize() before exiting a program. This routine 279*c4762a1bSJed Brown - finalizes the PETSc libraries as well as MPI 280*c4762a1bSJed Brown - provides summary and diagnostic information if certain runtime 281*c4762a1bSJed Brown options are chosen (e.g., -log_view). 282*c4762a1bSJed Brown */ 283*c4762a1bSJed Brown ierr = PetscFinalize(); 284*c4762a1bSJed Brown return ierr; 285*c4762a1bSJed Brown } 286*c4762a1bSJed Brown /* --------------------------------------------------------------------- */ 287*c4762a1bSJed Brown /* 288*c4762a1bSJed Brown InitialConditions - Computes the solution at the initial time. 289*c4762a1bSJed Brown 290*c4762a1bSJed Brown Input Parameter: 291*c4762a1bSJed Brown u - uninitialized solution vector (global) 292*c4762a1bSJed Brown appctx - user-defined application context 293*c4762a1bSJed Brown 294*c4762a1bSJed Brown Output Parameter: 295*c4762a1bSJed Brown u - vector with solution at initial time (global) 296*c4762a1bSJed Brown */ 297*c4762a1bSJed Brown PetscErrorCode InitialConditions(Vec u,AppCtx *appctx) 298*c4762a1bSJed Brown { 299*c4762a1bSJed Brown PetscScalar *u_localptr,h = appctx->h; 300*c4762a1bSJed Brown PetscInt i,mybase,myend; 301*c4762a1bSJed Brown PetscErrorCode ierr; 302*c4762a1bSJed Brown 303*c4762a1bSJed Brown /* 304*c4762a1bSJed Brown Determine starting point of each processor's range of 305*c4762a1bSJed Brown grid values. 306*c4762a1bSJed Brown */ 307*c4762a1bSJed Brown ierr = VecGetOwnershipRange(u,&mybase,&myend);CHKERRQ(ierr); 308*c4762a1bSJed Brown 309*c4762a1bSJed Brown /* 310*c4762a1bSJed Brown Get a pointer to vector data. 311*c4762a1bSJed Brown - For default PETSc vectors, VecGetArray() returns a pointer to 312*c4762a1bSJed Brown the data array. Otherwise, the routine is implementation dependent. 313*c4762a1bSJed Brown - You MUST call VecRestoreArray() when you no longer need access to 314*c4762a1bSJed Brown the array. 315*c4762a1bSJed Brown - Note that the Fortran interface to VecGetArray() differs from the 316*c4762a1bSJed Brown C version. See the users manual for details. 317*c4762a1bSJed Brown */ 318*c4762a1bSJed Brown ierr = VecGetArray(u,&u_localptr);CHKERRQ(ierr); 319*c4762a1bSJed Brown 320*c4762a1bSJed Brown /* 321*c4762a1bSJed Brown We initialize the solution array by simply writing the solution 322*c4762a1bSJed Brown directly into the array locations. Alternatively, we could use 323*c4762a1bSJed Brown VecSetValues() or VecSetValuesLocal(). 324*c4762a1bSJed Brown */ 325*c4762a1bSJed Brown for (i=mybase; i<myend; i++) u_localptr[i-mybase] = PetscSinScalar(PETSC_PI*i*6.*h) + 3.*PetscSinScalar(PETSC_PI*i*2.*h); 326*c4762a1bSJed Brown 327*c4762a1bSJed Brown /* 328*c4762a1bSJed Brown Restore vector 329*c4762a1bSJed Brown */ 330*c4762a1bSJed Brown ierr = VecRestoreArray(u,&u_localptr);CHKERRQ(ierr); 331*c4762a1bSJed Brown 332*c4762a1bSJed Brown /* 333*c4762a1bSJed Brown Print debugging information if desired 334*c4762a1bSJed Brown */ 335*c4762a1bSJed Brown if (appctx->debug) { 336*c4762a1bSJed Brown ierr = PetscPrintf(appctx->comm,"initial guess vector\n");CHKERRQ(ierr); 337*c4762a1bSJed Brown ierr = VecView(u,PETSC_VIEWER_STDOUT_WORLD);CHKERRQ(ierr); 338*c4762a1bSJed Brown } 339*c4762a1bSJed Brown 340*c4762a1bSJed Brown return 0; 341*c4762a1bSJed Brown } 342*c4762a1bSJed Brown /* --------------------------------------------------------------------- */ 343*c4762a1bSJed Brown /* 344*c4762a1bSJed Brown ExactSolution - Computes the exact solution at a given time. 345*c4762a1bSJed Brown 346*c4762a1bSJed Brown Input Parameters: 347*c4762a1bSJed Brown t - current time 348*c4762a1bSJed Brown solution - vector in which exact solution will be computed 349*c4762a1bSJed Brown appctx - user-defined application context 350*c4762a1bSJed Brown 351*c4762a1bSJed Brown Output Parameter: 352*c4762a1bSJed Brown solution - vector with the newly computed exact solution 353*c4762a1bSJed Brown */ 354*c4762a1bSJed Brown PetscErrorCode ExactSolution(PetscReal t,Vec solution,AppCtx *appctx) 355*c4762a1bSJed Brown { 356*c4762a1bSJed Brown PetscScalar *s_localptr,h = appctx->h,ex1,ex2,sc1,sc2; 357*c4762a1bSJed Brown PetscInt i,mybase,myend; 358*c4762a1bSJed Brown PetscErrorCode ierr; 359*c4762a1bSJed Brown 360*c4762a1bSJed Brown /* 361*c4762a1bSJed Brown Determine starting and ending points of each processor's 362*c4762a1bSJed Brown range of grid values 363*c4762a1bSJed Brown */ 364*c4762a1bSJed Brown ierr = VecGetOwnershipRange(solution,&mybase,&myend);CHKERRQ(ierr); 365*c4762a1bSJed Brown 366*c4762a1bSJed Brown /* 367*c4762a1bSJed Brown Get a pointer to vector data. 368*c4762a1bSJed Brown */ 369*c4762a1bSJed Brown ierr = VecGetArray(solution,&s_localptr);CHKERRQ(ierr); 370*c4762a1bSJed Brown 371*c4762a1bSJed Brown /* 372*c4762a1bSJed Brown Simply write the solution directly into the array locations. 373*c4762a1bSJed Brown Alternatively, we culd use VecSetValues() or VecSetValuesLocal(). 374*c4762a1bSJed Brown */ 375*c4762a1bSJed Brown ex1 = PetscExpReal(-36.*PETSC_PI*PETSC_PI*t); ex2 = PetscExpReal(-4.*PETSC_PI*PETSC_PI*t); 376*c4762a1bSJed Brown sc1 = PETSC_PI*6.*h; sc2 = PETSC_PI*2.*h; 377*c4762a1bSJed Brown for (i=mybase; i<myend; i++) s_localptr[i-mybase] = PetscSinScalar(sc1*(PetscReal)i)*ex1 + 3.*PetscSinScalar(sc2*(PetscReal)i)*ex2; 378*c4762a1bSJed Brown 379*c4762a1bSJed Brown /* 380*c4762a1bSJed Brown Restore vector 381*c4762a1bSJed Brown */ 382*c4762a1bSJed Brown ierr = VecRestoreArray(solution,&s_localptr);CHKERRQ(ierr); 383*c4762a1bSJed Brown return 0; 384*c4762a1bSJed Brown } 385*c4762a1bSJed Brown /* --------------------------------------------------------------------- */ 386*c4762a1bSJed Brown /* 387*c4762a1bSJed Brown Monitor - User-provided routine to monitor the solution computed at 388*c4762a1bSJed Brown each timestep. This example plots the solution and computes the 389*c4762a1bSJed Brown error in two different norms. 390*c4762a1bSJed Brown 391*c4762a1bSJed Brown Input Parameters: 392*c4762a1bSJed Brown ts - the timestep context 393*c4762a1bSJed Brown step - the count of the current step (with 0 meaning the 394*c4762a1bSJed Brown initial condition) 395*c4762a1bSJed Brown time - the current time 396*c4762a1bSJed Brown u - the solution at this timestep 397*c4762a1bSJed Brown ctx - the user-provided context for this monitoring routine. 398*c4762a1bSJed Brown In this case we use the application context which contains 399*c4762a1bSJed Brown information about the problem size, workspace and the exact 400*c4762a1bSJed Brown solution. 401*c4762a1bSJed Brown */ 402*c4762a1bSJed Brown PetscErrorCode Monitor(TS ts,PetscInt step,PetscReal time,Vec u,void *ctx) 403*c4762a1bSJed Brown { 404*c4762a1bSJed Brown AppCtx *appctx = (AppCtx*) ctx; /* user-defined application context */ 405*c4762a1bSJed Brown PetscErrorCode ierr; 406*c4762a1bSJed Brown PetscReal norm_2,norm_max; 407*c4762a1bSJed Brown 408*c4762a1bSJed Brown /* 409*c4762a1bSJed Brown View a graph of the current iterate 410*c4762a1bSJed Brown */ 411*c4762a1bSJed Brown ierr = VecView(u,appctx->viewer2);CHKERRQ(ierr); 412*c4762a1bSJed Brown 413*c4762a1bSJed Brown /* 414*c4762a1bSJed Brown Compute the exact solution 415*c4762a1bSJed Brown */ 416*c4762a1bSJed Brown ierr = ExactSolution(time,appctx->solution,appctx);CHKERRQ(ierr); 417*c4762a1bSJed Brown 418*c4762a1bSJed Brown /* 419*c4762a1bSJed Brown Print debugging information if desired 420*c4762a1bSJed Brown */ 421*c4762a1bSJed Brown if (appctx->debug) { 422*c4762a1bSJed Brown ierr = PetscPrintf(appctx->comm,"Computed solution vector\n");CHKERRQ(ierr); 423*c4762a1bSJed Brown ierr = VecView(u,PETSC_VIEWER_STDOUT_WORLD);CHKERRQ(ierr); 424*c4762a1bSJed Brown ierr = PetscPrintf(appctx->comm,"Exact solution vector\n");CHKERRQ(ierr); 425*c4762a1bSJed Brown ierr = VecView(appctx->solution,PETSC_VIEWER_STDOUT_WORLD);CHKERRQ(ierr); 426*c4762a1bSJed Brown } 427*c4762a1bSJed Brown 428*c4762a1bSJed Brown /* 429*c4762a1bSJed Brown Compute the 2-norm and max-norm of the error 430*c4762a1bSJed Brown */ 431*c4762a1bSJed Brown ierr = VecAXPY(appctx->solution,-1.0,u);CHKERRQ(ierr); 432*c4762a1bSJed Brown ierr = VecNorm(appctx->solution,NORM_2,&norm_2);CHKERRQ(ierr); 433*c4762a1bSJed Brown norm_2 = PetscSqrtReal(appctx->h)*norm_2; 434*c4762a1bSJed Brown ierr = VecNorm(appctx->solution,NORM_MAX,&norm_max);CHKERRQ(ierr); 435*c4762a1bSJed Brown if (norm_2 < 1e-14) norm_2 = 0; 436*c4762a1bSJed Brown if (norm_max < 1e-14) norm_max = 0; 437*c4762a1bSJed Brown 438*c4762a1bSJed Brown /* 439*c4762a1bSJed Brown PetscPrintf() causes only the first processor in this 440*c4762a1bSJed Brown communicator to print the timestep information. 441*c4762a1bSJed Brown */ 442*c4762a1bSJed Brown ierr = PetscPrintf(appctx->comm,"Timestep %D: time = %g 2-norm error = %g max norm error = %g\n",step,(double)time,(double)norm_2,(double)norm_max);CHKERRQ(ierr); 443*c4762a1bSJed Brown appctx->norm_2 += norm_2; 444*c4762a1bSJed Brown appctx->norm_max += norm_max; 445*c4762a1bSJed Brown 446*c4762a1bSJed Brown /* 447*c4762a1bSJed Brown View a graph of the error 448*c4762a1bSJed Brown */ 449*c4762a1bSJed Brown ierr = VecView(appctx->solution,appctx->viewer1);CHKERRQ(ierr); 450*c4762a1bSJed Brown 451*c4762a1bSJed Brown /* 452*c4762a1bSJed Brown Print debugging information if desired 453*c4762a1bSJed Brown */ 454*c4762a1bSJed Brown if (appctx->debug) { 455*c4762a1bSJed Brown ierr = PetscPrintf(appctx->comm,"Error vector\n");CHKERRQ(ierr); 456*c4762a1bSJed Brown ierr = VecView(appctx->solution,PETSC_VIEWER_STDOUT_WORLD);CHKERRQ(ierr); 457*c4762a1bSJed Brown } 458*c4762a1bSJed Brown 459*c4762a1bSJed Brown return 0; 460*c4762a1bSJed Brown } 461*c4762a1bSJed Brown 462*c4762a1bSJed Brown /* --------------------------------------------------------------------- */ 463*c4762a1bSJed Brown /* 464*c4762a1bSJed Brown RHSMatrixHeat - User-provided routine to compute the right-hand-side 465*c4762a1bSJed Brown matrix for the heat equation. 466*c4762a1bSJed Brown 467*c4762a1bSJed Brown Input Parameters: 468*c4762a1bSJed Brown ts - the TS context 469*c4762a1bSJed Brown t - current time 470*c4762a1bSJed Brown global_in - global input vector 471*c4762a1bSJed Brown dummy - optional user-defined context, as set by TSetRHSJacobian() 472*c4762a1bSJed Brown 473*c4762a1bSJed Brown Output Parameters: 474*c4762a1bSJed Brown AA - Jacobian matrix 475*c4762a1bSJed Brown BB - optionally different preconditioning matrix 476*c4762a1bSJed Brown str - flag indicating matrix structure 477*c4762a1bSJed Brown 478*c4762a1bSJed Brown Notes: 479*c4762a1bSJed Brown RHSMatrixHeat computes entries for the locally owned part of the system. 480*c4762a1bSJed Brown - Currently, all PETSc parallel matrix formats are partitioned by 481*c4762a1bSJed Brown contiguous chunks of rows across the processors. 482*c4762a1bSJed Brown - Each processor needs to insert only elements that it owns 483*c4762a1bSJed Brown locally (but any non-local elements will be sent to the 484*c4762a1bSJed Brown appropriate processor during matrix assembly). 485*c4762a1bSJed Brown - Always specify global row and columns of matrix entries when 486*c4762a1bSJed Brown using MatSetValues(); we could alternatively use MatSetValuesLocal(). 487*c4762a1bSJed Brown - Here, we set all entries for a particular row at once. 488*c4762a1bSJed Brown - Note that MatSetValues() uses 0-based row and column numbers 489*c4762a1bSJed Brown in Fortran as well as in C. 490*c4762a1bSJed Brown */ 491*c4762a1bSJed Brown PetscErrorCode RHSMatrixHeat(TS ts,PetscReal t,Vec X,Mat AA,Mat BB,void *ctx) 492*c4762a1bSJed Brown { 493*c4762a1bSJed Brown Mat A = AA; /* Jacobian matrix */ 494*c4762a1bSJed Brown AppCtx *appctx = (AppCtx*)ctx; /* user-defined application context */ 495*c4762a1bSJed Brown PetscErrorCode ierr; 496*c4762a1bSJed Brown PetscInt i,mstart,mend,idx[3]; 497*c4762a1bSJed Brown PetscScalar v[3],stwo = -2./(appctx->h*appctx->h),sone = -.5*stwo; 498*c4762a1bSJed Brown 499*c4762a1bSJed Brown /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 500*c4762a1bSJed Brown Compute entries for the locally owned part of the matrix 501*c4762a1bSJed Brown - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */ 502*c4762a1bSJed Brown 503*c4762a1bSJed Brown ierr = MatGetOwnershipRange(A,&mstart,&mend);CHKERRQ(ierr); 504*c4762a1bSJed Brown 505*c4762a1bSJed Brown /* 506*c4762a1bSJed Brown Set matrix rows corresponding to boundary data 507*c4762a1bSJed Brown */ 508*c4762a1bSJed Brown 509*c4762a1bSJed Brown if (mstart == 0) { /* first processor only */ 510*c4762a1bSJed Brown v[0] = 1.0; 511*c4762a1bSJed Brown ierr = MatSetValues(A,1,&mstart,1,&mstart,v,INSERT_VALUES);CHKERRQ(ierr); 512*c4762a1bSJed Brown mstart++; 513*c4762a1bSJed Brown } 514*c4762a1bSJed Brown 515*c4762a1bSJed Brown if (mend == appctx->m) { /* last processor only */ 516*c4762a1bSJed Brown mend--; 517*c4762a1bSJed Brown v[0] = 1.0; 518*c4762a1bSJed Brown ierr = MatSetValues(A,1,&mend,1,&mend,v,INSERT_VALUES);CHKERRQ(ierr); 519*c4762a1bSJed Brown } 520*c4762a1bSJed Brown 521*c4762a1bSJed Brown /* 522*c4762a1bSJed Brown Set matrix rows corresponding to interior data. We construct the 523*c4762a1bSJed Brown matrix one row at a time. 524*c4762a1bSJed Brown */ 525*c4762a1bSJed Brown v[0] = sone; v[1] = stwo; v[2] = sone; 526*c4762a1bSJed Brown for (i=mstart; i<mend; i++) { 527*c4762a1bSJed Brown idx[0] = i-1; idx[1] = i; idx[2] = i+1; 528*c4762a1bSJed Brown ierr = MatSetValues(A,1,&i,3,idx,v,INSERT_VALUES);CHKERRQ(ierr); 529*c4762a1bSJed Brown } 530*c4762a1bSJed Brown 531*c4762a1bSJed Brown /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 532*c4762a1bSJed Brown Complete the matrix assembly process and set some options 533*c4762a1bSJed Brown - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */ 534*c4762a1bSJed Brown /* 535*c4762a1bSJed Brown Assemble matrix, using the 2-step process: 536*c4762a1bSJed Brown MatAssemblyBegin(), MatAssemblyEnd() 537*c4762a1bSJed Brown Computations can be done while messages are in transition 538*c4762a1bSJed Brown by placing code between these two statements. 539*c4762a1bSJed Brown */ 540*c4762a1bSJed Brown ierr = MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 541*c4762a1bSJed Brown ierr = MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr); 542*c4762a1bSJed Brown 543*c4762a1bSJed Brown /* 544*c4762a1bSJed Brown Set and option to indicate that we will never add a new nonzero location 545*c4762a1bSJed Brown to the matrix. If we do, it will generate an error. 546*c4762a1bSJed Brown */ 547*c4762a1bSJed Brown ierr = MatSetOption(A,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);CHKERRQ(ierr); 548*c4762a1bSJed Brown 549*c4762a1bSJed Brown return 0; 550*c4762a1bSJed Brown } 551*c4762a1bSJed Brown 552*c4762a1bSJed Brown PetscErrorCode RHSFunctionHeat(TS ts,PetscReal t,Vec globalin,Vec globalout,void *ctx) 553*c4762a1bSJed Brown { 554*c4762a1bSJed Brown PetscErrorCode ierr; 555*c4762a1bSJed Brown Mat A; 556*c4762a1bSJed Brown 557*c4762a1bSJed Brown PetscFunctionBeginUser; 558*c4762a1bSJed Brown ierr = TSGetRHSJacobian(ts,&A,NULL,NULL,&ctx);CHKERRQ(ierr); 559*c4762a1bSJed Brown ierr = RHSMatrixHeat(ts,t,globalin,A,NULL,ctx);CHKERRQ(ierr); 560*c4762a1bSJed Brown /* ierr = MatView(A,PETSC_VIEWER_STDOUT_WORLD);CHKERRQ(ierr); */ 561*c4762a1bSJed Brown ierr = MatMult(A,globalin,globalout);CHKERRQ(ierr); 562*c4762a1bSJed Brown PetscFunctionReturn(0); 563*c4762a1bSJed Brown } 564*c4762a1bSJed Brown 565*c4762a1bSJed Brown /*TEST 566*c4762a1bSJed Brown 567*c4762a1bSJed Brown test: 568*c4762a1bSJed Brown args: -ts_view -nox 569*c4762a1bSJed Brown 570*c4762a1bSJed Brown test: 571*c4762a1bSJed Brown suffix: 2 572*c4762a1bSJed Brown args: -ts_view -nox 573*c4762a1bSJed Brown nsize: 3 574*c4762a1bSJed Brown 575*c4762a1bSJed Brown test: 576*c4762a1bSJed Brown suffix: 3 577*c4762a1bSJed Brown args: -ts_view -nox -nonlinear 578*c4762a1bSJed Brown 579*c4762a1bSJed Brown test: 580*c4762a1bSJed Brown suffix: 4 581*c4762a1bSJed Brown args: -ts_view -nox -nonlinear 582*c4762a1bSJed Brown nsize: 3 583*c4762a1bSJed Brown timeoutfactor: 3 584*c4762a1bSJed Brown 585*c4762a1bSJed Brown test: 586*c4762a1bSJed Brown suffix: sundials 587*c4762a1bSJed Brown requires: sundials 588*c4762a1bSJed Brown args: -nox -ts_type sundials -ts_max_steps 5 -nonlinear 589*c4762a1bSJed Brown nsize: 4 590*c4762a1bSJed Brown 591*c4762a1bSJed Brown TEST*/ 592*c4762a1bSJed Brown 593*c4762a1bSJed Brown 594