mirror of
https://git.dev.opencascade.org/repos/occt.git
synced 2025-04-16 10:08:36 +03:00
License statement text corrected; compiler warnings caused by Bison 2.41 disabled for MSVC; a few other compiler warnings on 54-bit Windows eliminated by appropriate type cast Wrong license statements corrected in several files. Copyright and license statements added in XSD and GLSL files. Copyright year updated in some files. Obsolete documentation files removed from DrawResources.
347 lines
12 KiB
C++
347 lines
12 KiB
C++
// Created on: 1997-05-28
|
|
// Created by: Sergey SOKOLOV
|
|
// Copyright (c) 1997-1999 Matra Datavision
|
|
// Copyright (c) 1999-2014 OPEN CASCADE SAS
|
|
//
|
|
// This file is part of Open CASCADE Technology software library.
|
|
//
|
|
// This library is free software; you can redistribute it and/or modify it under
|
|
// the terms of the GNU Lesser General Public License version 2.1 as published
|
|
// by the Free Software Foundation, with special exception defined in the file
|
|
// OCCT_LGPL_EXCEPTION.txt. Consult the file LICENSE_LGPL_21.txt included in OCCT
|
|
// distribution for complete text of the license and disclaimer of any warranty.
|
|
//
|
|
// Alternatively, this file may be used under the terms of Open CASCADE
|
|
// commercial license or contractual agreement.
|
|
|
|
#include <PLib_DoubleJacobiPolynomial.ixx>
|
|
#include <PLib_JacobiPolynomial.hxx>
|
|
#include <math_Vector.hxx>
|
|
|
|
//=======================================================================
|
|
//function : PLib_DoubleJacobiPolynomial
|
|
//purpose :
|
|
//=======================================================================
|
|
|
|
PLib_DoubleJacobiPolynomial::PLib_DoubleJacobiPolynomial()
|
|
|
|
{
|
|
}
|
|
|
|
//=======================================================================
|
|
//function : PLib_DoubleJacobiPolynomial
|
|
//purpose :
|
|
//=======================================================================
|
|
|
|
PLib_DoubleJacobiPolynomial::PLib_DoubleJacobiPolynomial(const Handle(PLib_JacobiPolynomial)& JacPolU,
|
|
const Handle(PLib_JacobiPolynomial)& JacPolV) :
|
|
myJacPolU(JacPolU),
|
|
myJacPolV(JacPolV)
|
|
{
|
|
Handle (TColStd_HArray1OfReal) TabMaxU =
|
|
new TColStd_HArray1OfReal (0,JacPolU->WorkDegree()-2*(JacPolU->NivConstr()+1));
|
|
JacPolU->MaxValue(TabMaxU->ChangeArray1());
|
|
myTabMaxU = TabMaxU;
|
|
|
|
Handle (TColStd_HArray1OfReal) TabMaxV =
|
|
new TColStd_HArray1OfReal (0,JacPolV->WorkDegree()-2*(JacPolV->NivConstr()+1));
|
|
JacPolV->MaxValue(TabMaxV->ChangeArray1());
|
|
myTabMaxV = TabMaxV;
|
|
}
|
|
|
|
//=======================================================================
|
|
//function : MaxErrorU
|
|
//purpose :
|
|
//=======================================================================
|
|
|
|
Standard_Real
|
|
PLib_DoubleJacobiPolynomial::MaxErrorU(const Standard_Integer Dimension,
|
|
const Standard_Integer DegreeU,
|
|
const Standard_Integer DegreeV,
|
|
const Standard_Integer dJacCoeff,
|
|
const TColStd_Array1OfReal& JacCoeff) const
|
|
{
|
|
Standard_Integer ii,idim,dJac,MinU,MinV,WorkDegreeU,WorkDegreeV;
|
|
Standard_Real Bid0;
|
|
|
|
math_Vector MaxErrDim(1,Dimension,0.);
|
|
|
|
MinU = 2*(myJacPolU->NivConstr()+1);
|
|
MinV = 2*(myJacPolV->NivConstr()+1);
|
|
WorkDegreeU = myJacPolU->WorkDegree();
|
|
WorkDegreeV = myJacPolV->WorkDegree();
|
|
|
|
Bid0 = myTabMaxV->Value(DegreeV-MinV);
|
|
for (idim=1; idim<=Dimension; idim++) {
|
|
dJac = dJacCoeff + (idim-1)*(WorkDegreeU+1)*(WorkDegreeV+1);
|
|
for (ii=MinU; ii<=DegreeU; ii++) {
|
|
MaxErrDim(idim) += (Abs(JacCoeff(ii + DegreeV*(WorkDegreeU+1) + dJac)) *
|
|
myTabMaxU->Value(ii-MinU) * Bid0);
|
|
}
|
|
}
|
|
return (MaxErrDim.Norm());
|
|
}
|
|
|
|
//=======================================================================
|
|
//function : MaxErrorV
|
|
//purpose :
|
|
//=======================================================================
|
|
|
|
Standard_Real
|
|
PLib_DoubleJacobiPolynomial::MaxErrorV(const Standard_Integer Dimension,
|
|
const Standard_Integer DegreeU,
|
|
const Standard_Integer DegreeV,
|
|
const Standard_Integer dJacCoeff,
|
|
const TColStd_Array1OfReal& JacCoeff) const
|
|
{
|
|
Standard_Integer jj,idim,dJac,MinU,MinV,WorkDegreeU,WorkDegreeV;
|
|
Standard_Real Bid0;
|
|
|
|
math_Vector MaxErrDim(1,Dimension,0.);
|
|
|
|
MinU = 2*(myJacPolU->NivConstr()+1);
|
|
MinV = 2*(myJacPolV->NivConstr()+1);
|
|
WorkDegreeU = myJacPolU->WorkDegree();
|
|
WorkDegreeV = myJacPolV->WorkDegree();
|
|
|
|
Bid0 = myTabMaxU->Value(DegreeU-MinU);
|
|
for (idim=1; idim<=Dimension; idim++) {
|
|
dJac = dJacCoeff + (idim-1)*(WorkDegreeU+1)*(WorkDegreeV+1);
|
|
for (jj=MinV; jj<=DegreeV; jj++) {
|
|
MaxErrDim(idim) += (Abs(JacCoeff(DegreeU + jj*(WorkDegreeU+1) + dJac)) *
|
|
myTabMaxV->Value(jj-MinV) * Bid0);
|
|
}
|
|
}
|
|
return (MaxErrDim.Norm());
|
|
}
|
|
|
|
//=======================================================================
|
|
//function : MaxError
|
|
//purpose :
|
|
//=======================================================================
|
|
|
|
Standard_Real
|
|
PLib_DoubleJacobiPolynomial::MaxError(const Standard_Integer Dimension,
|
|
const Standard_Integer MinDegreeU,
|
|
const Standard_Integer MaxDegreeU,
|
|
const Standard_Integer MinDegreeV,
|
|
const Standard_Integer MaxDegreeV,
|
|
const Standard_Integer dJacCoeff,
|
|
const TColStd_Array1OfReal& JacCoeff,
|
|
const Standard_Real Error) const
|
|
{
|
|
Standard_Integer ii,jj,idim,dJac,MinU,MinV,WorkDegreeU,WorkDegreeV;
|
|
Standard_Real Bid0,Bid1;
|
|
|
|
math_Vector MaxErrDim(1,Dimension,0.);
|
|
|
|
MinU = 2*(myJacPolU->NivConstr()+1);
|
|
MinV = 2*(myJacPolV->NivConstr()+1);
|
|
WorkDegreeU = myJacPolU->WorkDegree();
|
|
WorkDegreeV = myJacPolV->WorkDegree();
|
|
|
|
//------------------- Calcul du majorant de l'erreur max ---------------
|
|
//----- lorsque sont enleves les coeff. d'indices MinDegreeU a MaxDegreeU ------
|
|
//---------------- en U et d'indices MinDegreeV a MaxDegreeV en V --------------
|
|
|
|
for (idim=1; idim<=Dimension; idim++) {
|
|
dJac = dJacCoeff + (idim-1)*(WorkDegreeU+1)*(WorkDegreeV+1);
|
|
Bid1 = 0.;
|
|
for (jj=MinDegreeV; jj<=MaxDegreeV; jj++) {
|
|
Bid0 = 0.;
|
|
for (ii=MinDegreeU; ii<=MaxDegreeU; ii++) {
|
|
Bid0 += fabs(JacCoeff(ii + jj*(WorkDegreeU+1) + dJac)) * myTabMaxU->Value(ii-MinU);
|
|
}
|
|
Bid1 += Bid0 * myTabMaxV->Value(jj-MinV);
|
|
}
|
|
MaxErrDim(idim) = Bid1;
|
|
}
|
|
|
|
//----------------------- Calcul de l' erreur max ----------------------
|
|
|
|
math_Vector MaxErr2(1,2);
|
|
MaxErr2(1) = Error;
|
|
MaxErr2(2) = MaxErrDim.Norm();
|
|
return (MaxErr2.Norm());
|
|
}
|
|
|
|
//=======================================================================
|
|
//function : ReduceDegree
|
|
//purpose :
|
|
//=======================================================================
|
|
|
|
void PLib_DoubleJacobiPolynomial::ReduceDegree(const Standard_Integer Dimension,
|
|
const Standard_Integer MinDegreeU,
|
|
const Standard_Integer MaxDegreeU,
|
|
const Standard_Integer MinDegreeV,
|
|
const Standard_Integer MaxDegreeV,
|
|
const Standard_Integer dJacCoeff,
|
|
const TColStd_Array1OfReal& JacCoeff,
|
|
const Standard_Real EpmsCut,
|
|
Standard_Real& MaxError,
|
|
Standard_Integer& NewDegreeU,
|
|
Standard_Integer& NewDegreeV) const
|
|
{
|
|
Standard_Integer NewU,NewV;
|
|
Standard_Real ErrU,ErrV;
|
|
|
|
NewU = MaxDegreeU;
|
|
NewV = MaxDegreeV;
|
|
math_Vector MaxErr2(1,2);
|
|
|
|
//**********************************************************************
|
|
//-------------------- Coupure des coefficients ------------------------
|
|
//**********************************************************************
|
|
|
|
do {
|
|
|
|
//------------------- Calcul du majorant de l'erreur max ---------------
|
|
//----- lorsque sont enleves les coeff. d'indices MinU a NewU ------
|
|
//---------------- en U, le degre en V etant fixe a NewV -----------------
|
|
if (NewV > MinDegreeV)
|
|
ErrV = MaxErrorU(Dimension,NewU,NewV,dJacCoeff,JacCoeff);
|
|
else {
|
|
ErrV = 2*EpmsCut;
|
|
}
|
|
|
|
//------------------- Calcul du majorant de l'erreur max ---------------
|
|
//----- lorsque sont enleves les coeff. d'indices MinV a NewV ------
|
|
//---------------- en V, le degre en U etant fixe a NewU -----------------
|
|
if (NewU > MinDegreeU)
|
|
ErrU = MaxErrorV(Dimension,NewU,NewV,dJacCoeff,JacCoeff);
|
|
else {
|
|
ErrU = 2*EpmsCut;
|
|
}
|
|
|
|
//----------------------- Calcul de l' erreur max ----------------------
|
|
MaxErr2(1) = MaxError;
|
|
MaxErr2(2) = ErrU;
|
|
ErrU = MaxErr2.Norm();
|
|
MaxErr2(2) = ErrV;
|
|
ErrV = MaxErr2.Norm();
|
|
|
|
if (ErrU > ErrV) {
|
|
if (ErrV < EpmsCut) {
|
|
MaxError = ErrV;
|
|
NewV--;
|
|
}
|
|
}
|
|
else {
|
|
if (ErrU < EpmsCut) {
|
|
MaxError = ErrU;
|
|
NewU--;
|
|
}
|
|
}
|
|
}
|
|
while ((ErrU > ErrV && ErrV <= EpmsCut) || (ErrV >= ErrU && ErrU <= EpmsCut));
|
|
|
|
//-------------------------- Recuperation des degres -------------------
|
|
|
|
NewDegreeU = Max(NewU,1);
|
|
NewDegreeV = Max(NewV,1);
|
|
}
|
|
|
|
//=======================================================================
|
|
//function : AverageError
|
|
//purpose :
|
|
//=======================================================================
|
|
|
|
Standard_Real
|
|
PLib_DoubleJacobiPolynomial::AverageError(const Standard_Integer Dimension,
|
|
const Standard_Integer DegreeU,
|
|
const Standard_Integer DegreeV,
|
|
const Standard_Integer dJacCoeff,
|
|
const TColStd_Array1OfReal& JacCoeff) const
|
|
{
|
|
Standard_Integer ii,jj,idim,dJac,IDebU,IDebV,MinU,MinV,WorkDegreeU,WorkDegreeV;
|
|
Standard_Real Bid0,Bid1,AverageErr;
|
|
|
|
//----------------------------- Initialisations ------------------------
|
|
|
|
IDebU = 2*(myJacPolU->NivConstr()+1);
|
|
IDebV = 2*(myJacPolV->NivConstr()+1);
|
|
MinU = Max(IDebU,DegreeU);
|
|
MinV = Max(IDebV,DegreeV);
|
|
WorkDegreeU = myJacPolU->WorkDegree();
|
|
WorkDegreeV = myJacPolV->WorkDegree();
|
|
Bid0 = 0.;
|
|
|
|
//------------------ Calcul du majorant de l'erreur moyenne ------------
|
|
//----- lorsque sont enleves les coeff. d'indices DegreeU a WorkDegreeU ------
|
|
//---------------- en U et d'indices DegreeV a WorkDegreeV en V --------------
|
|
|
|
for (idim=1; idim<=Dimension; idim++) {
|
|
dJac = dJacCoeff + (idim-1)*(WorkDegreeU+1)*(WorkDegreeV+1);
|
|
for (jj=MinV; jj<=WorkDegreeV; jj++) {
|
|
for (ii=IDebU; ii<=WorkDegreeU; ii++) {
|
|
Bid1 = JacCoeff(ii + jj*(WorkDegreeU+1) + dJac);
|
|
Bid0 += Bid1*Bid1;
|
|
}
|
|
}
|
|
for (jj=IDebV; jj<=MinV-1; jj++) {
|
|
for (ii=MinU; ii<=WorkDegreeU; ii++) {
|
|
Bid1 = JacCoeff(ii + jj*(WorkDegreeU+1) + dJac);
|
|
Bid0 += Bid1*Bid1;
|
|
}
|
|
}
|
|
}
|
|
AverageErr = sqrt(Bid0/4);
|
|
return (AverageErr);
|
|
}
|
|
|
|
//=======================================================================
|
|
//function : WDoubleJacobiToCoefficients
|
|
//purpose :
|
|
//=======================================================================
|
|
|
|
void PLib_DoubleJacobiPolynomial::WDoubleJacobiToCoefficients(const Standard_Integer Dimension,
|
|
const Standard_Integer DegreeU,
|
|
const Standard_Integer DegreeV,
|
|
const TColStd_Array1OfReal& JacCoeff,
|
|
TColStd_Array1OfReal& Coefficients) const
|
|
{
|
|
Standard_Integer iu,iv,idim,WorkDegreeU,WorkDegreeV;
|
|
|
|
Coefficients.Init(0.);
|
|
|
|
WorkDegreeU = myJacPolU->WorkDegree();
|
|
WorkDegreeV = myJacPolV->WorkDegree();
|
|
|
|
TColStd_Array1OfReal Aux1(0, (DegreeU+1)*(DegreeV+1)*Dimension-1);
|
|
TColStd_Array1OfReal Aux2(0, (DegreeU+1)*(DegreeV+1)*Dimension-1);
|
|
|
|
for (iu=0; iu<=DegreeU; iu++) {
|
|
for (iv=0; iv<=DegreeV; iv++) {
|
|
for (idim=1; idim<=Dimension; idim++) {
|
|
Aux1(idim-1 + iv*Dimension + iu*Dimension*(DegreeV+1)) =
|
|
JacCoeff(iu + iv*(WorkDegreeU+1) + (idim-1)*(WorkDegreeU+1)*(WorkDegreeV+1));
|
|
}
|
|
}
|
|
}
|
|
// Passage dans canonique en u.
|
|
myJacPolU->ToCoefficients(Dimension*(DegreeV+1),DegreeU,Aux1,Aux2);
|
|
|
|
// Permutation des u et des v.
|
|
for (iu=0; iu<=DegreeU; iu++) {
|
|
for (iv=0; iv<=DegreeV; iv++) {
|
|
for (idim=1; idim<=Dimension; idim++) {
|
|
Aux1(idim-1 + iu*Dimension + iv*Dimension*(DegreeU+1)) =
|
|
Aux2(idim-1 + iv*Dimension + iu*Dimension*(DegreeV+1));
|
|
}
|
|
}
|
|
}
|
|
// Passage dans canonique en v.
|
|
myJacPolV->ToCoefficients(Dimension*(DegreeU+1),DegreeV,Aux1,Aux2);
|
|
|
|
// Permutation des u et des v.
|
|
for (iu=0; iu<=DegreeU; iu++) {
|
|
for (iv=0; iv<=DegreeV; iv++) {
|
|
for (idim=1; idim<=Dimension; idim++) {
|
|
Coefficients(iu + iv*(DegreeU+1) + (idim-1)*(DegreeU+1)*(DegreeV+1)) =
|
|
Aux2(idim-1 + iu*Dimension + iv*Dimension*(DegreeU+1));
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|