int LOCATest(MFNVector u, void *d, MFErrorHandler err)
{
LOCANVectorData* v_data = (LOCANVectorData *)MFNVectorGetData(u,err);
LOCAData* data = (LOCAData*) d;
list<ParamData>::iterator it = data->paramData->begin();
for (unsigned int i=0; i<data->paramData->size(); i++) {
if (v_data->u_ptr->getScalar(i) < it->minValue)
return 0;
if (v_data->u_ptr->getScalar(i) > it->maxValue)
return 0;
++it;
}
if (v_data->u_ptr->getXVec()->norm(NOX::Abstract::Vector::MaxNorm) >
data->solutionMax)
return 0;
return 1;
}
double *MFNV_CStar(MFNVector thisVector, MFErrorHandler e)
{
static char RoutineName[]={"MFNV_CStar"};
struct MFDenseNVectorData *u;
#ifdef MFNOCONFIDENCE
if(thisVector==NULL)
{
sprintf(MFNVectorErrorMsg,"Pointer to Vector (argument 1) is NULL");
MFSetError(e,4,RoutineName,MFNVectorErrorMsg,__LINE__,__FILE__);
return NULL;
}
if(strcmp(MFNVGetId(thisVector,e),"DENSE"))
{
sprintf(MFNVectorErrorMsg,"Trying to get dense arrays from non-dense Vector type \"%s\"",MFNVGetId(thisVector,e));
MFSetError(e,12,RoutineName,MFNVectorErrorMsg,__LINE__,__FILE__);
return NULL;
}
#endif
u=(struct MFDenseNVectorData*)MFNVectorGetData(thisVector,e);
return u->data;
}
int MFLOCAProjectToDraw(MFNVector u, double *x, void *d, MFErrorHandler err)
{
LOCAData* data = (LOCAData *)d;
if(x==(double*)NULL)
return data->grp->projectToDrawDimension();
LOCANVectorData* v_data = (LOCANVectorData *)MFNVectorGetData(u, err);
data->grp->projectToDraw(*(v_data->u_ptr), x);
return 0;
}
double MFScaleLOCA(int n,int k,MFNVector u,MFNKMatrix Phi,void *d,
MFErrorHandler err)
{
LOCAData* data = (LOCAData *)d;
if (data->radius < 0.0) {
data->radius = 0.0;
data->minRadius = 0.0;
data->maxRadius = 0.0;
std::list<ParamData>::iterator it = data->paramData->begin();
for (int i=0; i<k; i++) {
double dpidsj = 0.0;
for (int j=0; j<k; j++) {
MFNVector tmp = MFMColumn(Phi,j, err);
LOCANVectorData* tmp2_data = (LOCANVectorData *) MFNVectorGetData(tmp, err);
dpidsj += fabs(tmp2_data->u_ptr->getScalar(i));
MFFreeNVector(tmp, err);
}
data->radius += it->initialStepSize / dpidsj;
data->minRadius += it->minStepSize / dpidsj;
data->maxRadius += it->maxStepSize / dpidsj;
it++;
}
data->radius /= (double) k;
data->minRadius /= (double) k;
data->maxRadius /= (double) k;
}
else {
NOX::StatusTest::StatusType status = data->solver->getStatus();
if (status != NOX::StatusTest::Converged) {
data->radius *= 0.7;
if (data->radius < data->minRadius)
data->globalData->locaErrorCheck->throwError(
"MFScaleLOCA()",
"Reached minimum radius!");
}
else {
double numNonlinearSteps =
static_cast<double>(data->solver->getNumIterations());
double factor = (data->maxNonlinearIterations - numNonlinearSteps)
/ (data->maxNonlinearIterations);
data->radius *= (1.0 + data->aggressiveness * factor * factor);
if (data->radius > data->maxRadius)
data->radius = data->maxRadius;
}
}
data->globalData->locaUtils->out()
<< "radius = " << data->radius << std::endl;
return data->radius;
}
int MFTangentLOCA(int n,int k,MFNVector vu,MFNKMatrix mPhi,void *d,
MFErrorHandler err)
{
LOCAData* data = (LOCAData *)d;
LOCANVectorData* u0_data = (LOCANVectorData *) MFNVectorGetData(vu, err);
data->grp->setX(*(u0_data->u_ptr));
data->grp->computePredictor();
const LOCA::MultiContinuation::ExtendedMultiVector& pred =
data->grp->getPredictorTangent();
for (int i=0; i<k; i++) {
Teuchos::RCP<LMCEV> t =
Teuchos::rcp_dynamic_cast<LMCEV>(pred[i].clone());
MFNVector tmp = MFCreateLOCANVectorWithData(t,err);
MFMSetColumn(mPhi, i, tmp, err);
MFFreeNVector(tmp, err);
}
MFGramSchmidt(data->space,mPhi, err);
return 1;
}
int MFProjectLOCA(int n,int k,MFNVector vu0,MFNKMatrix mPhi,MFNVector vu,
void *d,int *index, MFErrorHandler err)
{
static int stepNumber = 1;
int i;
*index = 1;
LOCAData* data = (LOCAData *)d;
for (i=0; i<k; i++) {
MFNVector tmp = MFMColumn(mPhi,i, err);
LOCANVectorData* tmp2_data = (LOCANVectorData *) MFNVectorGetData(tmp, err);
data->grp->setPredictorTangentDirection(*(tmp2_data->u_ptr), i);
MFFreeNVector(tmp, err);
}
LOCANVectorData* u0_data = (LOCANVectorData *) MFNVectorGetData(vu0, err);
data->grp->setPrevX(*(u0_data->u_ptr));
data->grp->setX(*(u0_data->u_ptr));
for (i=0; i<k; i++) {
data->grp->setStepSize(0.0, i);
}
if (data->globalData->locaUtils->isPrintType(NOX::Utils::StepperIteration)) {
data->globalData->locaUtils->out()
<< "\n" << data->globalData->locaUtils->fill(72, '~') << "\n";
data->globalData->locaUtils->out()
<< "Start of Continuation Step " << stepNumber <<" : " << std::endl;
std::list<ParamData>::iterator it = data->paramData->begin();
for (i=0; i<k; i++) {
data->globalData->locaUtils->out()
<< "\tParameter: " << it->name << " = "
<< data->globalData->locaUtils->sciformat(data->grp->getContinuationParameter(i))
<< std::endl;
it++;
}
data->globalData->locaUtils->out()
<< data->globalData->locaUtils->fill(72, '~') << "\n" << std::endl;
}
data->grp->preProcessContinuationStep(LOCA::Abstract::Iterator::Successful);
data->grp->computeF();
data->solver->reset(data->grp->getX());
NOX::StatusTest::StatusType status = data->solver->solve();
if (status != NOX::StatusTest::Converged) {
if (data->globalData->locaUtils->isPrintType(NOX::Utils::StepperIteration)) {
data->globalData->locaUtils->out()
<< std::endl << data->globalData->locaUtils->fill(72, '~')
<< std::endl;
data->globalData->locaUtils->out()
<< "Continuation Step Number " << stepNumber
<< " experienced a convergence failure in\n"
<< "the nonlinear solver after "<< data->solver->getNumIterations()
<<" Iterations\n";
data->globalData->locaUtils->out()
<< "Value of continuation parameters at failed step:" << std::endl;
std::list<ParamData>::iterator it = data->paramData->begin();
for (i=0; i<k; i++) {
data->globalData->locaUtils->out()
<< "\tParameter: " << it->name << " = "
<< data->globalData->locaUtils->sciformat(data->grp->getContinuationParameter(i))
<< std::endl;
it++;
}
data->globalData->locaUtils->out()
<< data->globalData->locaUtils->fill(72, '~') << std::endl;
}
data->grp->postProcessContinuationStep(LOCA::Abstract::Iterator::Unsuccessful);
return 0;
}
else {
LOCANVectorData* u_data = (LOCANVectorData *) MFNVectorGetData(vu, err);
*(Teuchos::rcp_dynamic_cast<NOX::Abstract::Group>(data->grp)) =
data->solver->getSolutionGroup();
*(u_data->u_ptr) = data->grp->getX(); /* overloaded deep copy */
data->grp->postProcessContinuationStep(LOCA::Abstract::Iterator::Successful);
if (data->globalData->locaUtils->isPrintType(NOX::Utils::StepperIteration)) {
data->globalData->locaUtils->out()
<< "\n" << data->globalData->locaUtils->fill(72, '~') << "\n";
data->globalData->locaUtils->out()
<< "End of Continuation Step " << stepNumber << " : " << std::endl;
std::list<ParamData>::iterator it = data->paramData->begin();
for (i=0; i<k; i++) {
data->globalData->locaUtils->out()
<< "\tParameter: " << it->name << " = "
<< data->globalData->locaUtils->sciformat(data->grp->getContinuationParameter(i))
<< std::endl;
it++;
}
data->globalData->locaUtils->out()
<< "--> Step Converged in "
<< data->solver->getNumIterations()
<<" Nonlinear Solver Iterations!\n";
data->globalData->locaUtils->out()
<< data->globalData->locaUtils->fill(72, '~') << "\n" << std::endl;
}
++stepNumber;
return 1;
}
}
