Actual source code: mfnsolve.c

slepc-3.11.2 2019-07-30
Report Typos and Errors
  1: /*
  2:    - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
  3:    SLEPc - Scalable Library for Eigenvalue Problem Computations
  4:    Copyright (c) 2002-2019, Universitat Politecnica de Valencia, Spain

  6:    This file is part of SLEPc.
  7:    SLEPc is distributed under a 2-clause BSD license (see LICENSE).
  8:    - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
  9: */
 10: /*
 11:    MFN routines related to the solution process
 12: */

 14: #include <slepc/private/mfnimpl.h>   /*I "slepcmfn.h" I*/

 16: /*@
 17:    MFNSolve - Solves the matrix function problem. Given a vector b, the
 18:    vector x = f(A)*b is returned.

 20:    Collective on MFN

 22:    Input Parameters:
 23: +  mfn - matrix function context obtained from MFNCreate()
 24: -  b   - the right hand side vector

 26:    Output Parameter:
 27: .  x   - the solution (this may be the same vector as b, then b will be
 28:          overwritten with the answer)

 30:    Options Database Keys:
 31: +  -mfn_view - print information about the solver used
 32: .  -mfn_view_mat binary - save the matrix to the default binary viewer
 33: .  -mfn_view_rhs binary - save right hand side vector to the default binary viewer
 34: .  -mfn_view_solution binary - save computed solution vector to the default binary viewer
 35: -  -mfn_converged_reason - print reason for convergence, and number of iterations

 37:    Notes:
 38:    The matrix A is specified with MFNSetOperator().
 39:    The function f is specified with MFNSetFN().

 41:    Level: beginner

 43: .seealso: MFNCreate(), MFNSetUp(), MFNDestroy(), MFNSetTolerances(),
 44:           MFNSetOperator(), MFNSetFN()
 45: @*/
 46: PetscErrorCode MFNSolve(MFN mfn,Vec b,Vec x)
 47: {

 56:   VecSetErrorIfLocked(x,3);

 58:   /* call setup */
 59:   MFNSetUp(mfn);
 60:   mfn->its = 0;

 62:   MFNViewFromOptions(mfn,NULL,"-mfn_view_pre");

 64:   /* check nonzero right-hand side */
 65:   VecNorm(b,NORM_2,&mfn->bnorm);
 66:   if (!mfn->bnorm) SETERRQ(PetscObjectComm((PetscObject)mfn),PETSC_ERR_ARG_WRONG,"Cannot pass a zero b vector to MFNSolve()");

 68:   /* call solver */
 69:   PetscLogEventBegin(MFN_Solve,mfn,b,x,0);
 70:   if (b!=x) { VecLockReadPush(b); }
 71:   (*mfn->ops->solve)(mfn,b,x);
 72:   if (b!=x) { VecLockReadPop(b); }
 73:   PetscLogEventEnd(MFN_Solve,mfn,b,x,0);

 75:   if (!mfn->reason) SETERRQ(PetscObjectComm((PetscObject)mfn),PETSC_ERR_PLIB,"Internal error, solver returned without setting converged reason");

 77:   if (mfn->errorifnotconverged && mfn->reason < 0) SETERRQ(PetscObjectComm((PetscObject)mfn),PETSC_ERR_NOT_CONVERGED,"MFNSolve has not converged");

 79:   /* various viewers */
 80:   MFNViewFromOptions(mfn,NULL,"-mfn_view");
 81:   MFNReasonViewFromOptions(mfn);
 82:   MatViewFromOptions(mfn->A,(PetscObject)mfn,"-mfn_view_mat");
 83:   VecViewFromOptions(b,(PetscObject)mfn,"-mfn_view_rhs");
 84:   VecViewFromOptions(x,(PetscObject)mfn,"-mfn_view_solution");
 85:   return(0);
 86: }

 88: /*@
 89:    MFNGetIterationNumber - Gets the current iteration number. If the
 90:    call to MFNSolve() is complete, then it returns the number of iterations
 91:    carried out by the solution method.

 93:    Not Collective

 95:    Input Parameter:
 96: .  mfn - the matrix function context

 98:    Output Parameter:
 99: .  its - number of iterations

101:    Note:
102:    During the i-th iteration this call returns i-1. If MFNSolve() is
103:    complete, then parameter "its" contains either the iteration number at
104:    which convergence was successfully reached, or failure was detected.
105:    Call MFNGetConvergedReason() to determine if the solver converged or
106:    failed and why.

108:    Level: intermediate

110: .seealso: MFNGetConvergedReason(), MFNSetTolerances()
111: @*/
112: PetscErrorCode MFNGetIterationNumber(MFN mfn,PetscInt *its)
113: {
117:   *its = mfn->its;
118:   return(0);
119: }

121: /*@
122:    MFNGetConvergedReason - Gets the reason why the MFNSolve() iteration was
123:    stopped.

125:    Not Collective

127:    Input Parameter:
128: .  mfn - the matrix function context

130:    Output Parameter:
131: .  reason - negative value indicates diverged, positive value converged

133:    Notes:

135:    Possible values for reason are
136: +  MFN_CONVERGED_TOL - converged up to tolerance
137: .  MFN_CONVERGED_ITS - solver completed the requested number of steps
138: .  MFN_DIVERGED_ITS - required more than max_it iterations to reach convergence
139: -  MFN_DIVERGED_BREAKDOWN - generic breakdown in method

141:    Can only be called after the call to MFNSolve() is complete.

143:    Basic solvers (e.g. unrestarted Krylov iterations) cannot determine if the
144:    computation is accurate up to the requested tolerance. In that case, the
145:    converged reason is set to MFN_CONVERGED_ITS if the requested number of steps
146:    (for instance, the ncv value in unrestarted Krylov methods) have been
147:    completed successfully.

149:    Level: intermediate

151: .seealso: MFNSetTolerances(), MFNSolve(), MFNConvergedReason, MFNSetErrorIfNotConverged()
152: @*/
153: PetscErrorCode MFNGetConvergedReason(MFN mfn,MFNConvergedReason *reason)
154: {
158:   *reason = mfn->reason;
159:   return(0);
160: }