Actual source code: shift.c

slepc-3.7.1 2016-05-27
Report Typos and Errors
  1: /*
  2:     Shift spectral transformation, applies (A + sigma I) as operator, or
  3:     inv(B)(A + sigma B) for generalized problems

  5:    - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
  6:    SLEPc - Scalable Library for Eigenvalue Problem Computations
  7:    Copyright (c) 2002-2016, Universitat Politecnica de Valencia, Spain

  9:    This file is part of SLEPc.

 11:    SLEPc is free software: you can redistribute it and/or modify it under  the
 12:    terms of version 3 of the GNU Lesser General Public License as published by
 13:    the Free Software Foundation.

 15:    SLEPc  is  distributed in the hope that it will be useful, but WITHOUT  ANY
 16:    WARRANTY;  without even the implied warranty of MERCHANTABILITY or  FITNESS
 17:    FOR  A  PARTICULAR PURPOSE. See the GNU Lesser General Public  License  for
 18:    more details.

 20:    You  should have received a copy of the GNU Lesser General  Public  License
 21:    along with SLEPc. If not, see <http://www.gnu.org/licenses/>.
 22:    - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
 23: */

 25: #include <slepc/private/stimpl.h>

 29: PetscErrorCode STApply_Shift(ST st,Vec x,Vec y)
 30: {

 34:   if (st->nmat>1) {
 35:     /* generalized eigenproblem: y = B^-1 (A - sB) x */
 36:     MatMult(st->T[0],x,st->w);
 37:     STMatSolve(st,st->w,y);
 38:   } else {
 39:     /* standard eigenproblem: y = (A - sI) x */
 40:     MatMult(st->T[0],x,y);
 41:   }
 42:   return(0);
 43: }

 47: PetscErrorCode STApplyTranspose_Shift(ST st,Vec x,Vec y)
 48: {

 52:   if (st->nmat>1) {
 53:     /* generalized eigenproblem: y = (A - sB)^T B^-T  x */
 54:     STMatSolveTranspose(st,x,st->w);
 55:     MatMultTranspose(st->T[0],st->w,y);
 56:   } else {
 57:     /* standard eigenproblem: y = (A^T - sI) x */
 58:     MatMultTranspose(st->T[0],x,y);
 59:   }
 60:   return(0);
 61: }

 65: PetscErrorCode STBackTransform_Shift(ST st,PetscInt n,PetscScalar *eigr,PetscScalar *eigi)
 66: {
 67:   PetscInt j;

 70:   for (j=0;j<n;j++) {
 71:     eigr[j] += st->sigma;
 72:   }
 73:   return(0);
 74: }

 78: PetscErrorCode STPostSolve_Shift(ST st)
 79: {

 83:   if (st->shift_matrix == ST_MATMODE_INPLACE) {
 84:     if (st->nmat>1) {
 85:       MatAXPY(st->A[0],st->sigma,st->A[1],st->str);
 86:     } else {
 87:       MatShift(st->A[0],st->sigma);
 88:     }
 89:     st->Astate[0] = ((PetscObject)st->A[0])->state;
 90:     st->state = ST_STATE_INITIAL;
 91:   }
 92:   return(0);
 93: }

 97: PetscErrorCode STSetUp_Shift(ST st)
 98: {
100:   PetscInt       k,nc,nmat=PetscMax(st->nmat,2);
101:   PetscScalar    *coeffs=NULL;

104:   if (nmat<3 || st->transform) {
105:     if (nmat>2) {
106:       nc = (nmat*(nmat+1))/2;
107:       PetscMalloc(nc*sizeof(PetscScalar),&coeffs);
108:       /* Compute coeffs */
109:       STCoeffs_Monomial(st,coeffs);
110:     }
111:     /* T[n] = A_n */
112:     k = nmat-1;
113:     PetscObjectReference((PetscObject)st->A[k]);
114:     MatDestroy(&st->T[k]);
115:     st->T[k] = st->A[k];
116:     for (k=0;k<nmat-1;k++) {
117:       STMatMAXPY_Private(st,nmat>2?st->sigma:-st->sigma,0.0,k,coeffs?coeffs+((nmat-k)*(nmat-k-1))/2:NULL,PetscNot(st->state==ST_STATE_UPDATED),&st->T[k]);
118:     }
119:      if (nmat>2) { PetscFree(coeffs); }
120:   } else {
121:     for (k=0;k<nmat;k++) {
122:       PetscObjectReference((PetscObject)st->A[k]);
123:       MatDestroy(&st->T[k]);
124:       st->T[k] = st->A[k];
125:     }
126:   }
127:   if (nmat>=2 && st->transform) {
128:     PetscObjectReference((PetscObject)st->T[nmat-1]);
129:     MatDestroy(&st->P);
130:     st->P = st->T[nmat-1];
131:   }
132:   if (st->P) {
133:     if (!st->ksp) { STGetKSP(st,&st->ksp); }
134:     STCheckFactorPackage(st);
135:     KSPSetOperators(st->ksp,st->P,st->P);
136:     KSPSetErrorIfNotConverged(st->ksp,PETSC_TRUE);
137:     KSPSetUp(st->ksp);
138:   }
139:   return(0);
140: }

144: PetscErrorCode STSetShift_Shift(ST st,PetscScalar newshift)
145: {
147:   PetscInt       k,nc,nmat=PetscMax(st->nmat,2);
148:   PetscScalar    *coeffs=NULL;

151:   if (st->transform) {
152:     if (st->shift_matrix == ST_MATMODE_COPY && nmat>2) {
153:       nc = (nmat*(nmat+1))/2;
154:       PetscMalloc(nc*sizeof(PetscScalar),&coeffs);
155:       /* Compute coeffs */
156:       STCoeffs_Monomial(st,coeffs);
157:     }
158:     for (k=0;k<nmat-1;k++) {
159:       STMatMAXPY_Private(st,nmat>2?newshift:-newshift,nmat>2?st->sigma:-st->sigma,k,coeffs?coeffs+((nmat-k)*(nmat-k-1))/2:NULL,PETSC_FALSE,&st->T[k]);
160:     }
161:     if (st->shift_matrix == ST_MATMODE_COPY && nmat>2) {
162:         PetscFree(coeffs);
163:     }
164:   }
165:   return(0);
166: }

170: PetscErrorCode STSetFromOptions_Shift(PetscOptionItems *PetscOptionsObject,ST st)
171: {
173:   PC             pc;
174:   PCType         pctype;
175:   KSPType        ksptype;

178:   if (!st->ksp) { STGetKSP(st,&st->ksp); }
179:   KSPGetPC(st->ksp,&pc);
180:   KSPGetType(st->ksp,&ksptype);
181:   PCGetType(pc,&pctype);
182:   if (!pctype && !ksptype) {
183:     if (st->shift_matrix == ST_MATMODE_SHELL) {
184:       /* in shell mode use GMRES with Jacobi as the default */
185:       KSPSetType(st->ksp,KSPGMRES);
186:       PCSetType(pc,PCJACOBI);
187:     } else {
188:       /* use direct solver as default */
189:       KSPSetType(st->ksp,KSPPREONLY);
190:       PCSetType(pc,PCLU);
191:     }
192:   }
193:   return(0);
194: }

198: PETSC_EXTERN PetscErrorCode STCreate_Shift(ST st)
199: {
201:   st->ops->apply           = STApply_Shift;
202:   st->ops->getbilinearform = STGetBilinearForm_Default;
203:   st->ops->applytrans      = STApplyTranspose_Shift;
204:   st->ops->postsolve       = STPostSolve_Shift;
205:   st->ops->backtransform   = STBackTransform_Shift;
206:   st->ops->setfromoptions  = STSetFromOptions_Shift;
207:   st->ops->setup           = STSetUp_Shift;
208:   st->ops->setshift        = STSetShift_Shift;
209:   return(0);
210: }