42 #ifndef _TEUCHOS_SERIALBANDDENSEMATRIX_HPP_
43 #define _TEUCHOS_SERIALBANDDENSEMATRIX_HPP_
54 #include "Teuchos_Assert.hpp"
132 template<
typename OrdinalType,
typename ScalarType>
293 ScalarType&
operator () (OrdinalType rowIndex, OrdinalType colIndex);
303 const ScalarType&
operator () (OrdinalType rowIndex, OrdinalType colIndex)
const;
323 const ScalarType*
operator [] (OrdinalType colIndex)
const;
327 ScalarType*
values()
const {
return(values_); }
357 int scale (
const ScalarType alpha );
391 OrdinalType
numRows()
const {
return(numRows_); }
394 OrdinalType
numCols()
const {
return(numCols_); }
403 OrdinalType
stride()
const {
return(stride_); }
406 bool empty()
const {
return(numRows_ == 0 || numCols_ == 0); }
424 virtual void print(std::ostream& os)
const;
429 void copyMat(ScalarType* inputMatrix, OrdinalType strideInput,
430 OrdinalType
numRows, OrdinalType
numCols, ScalarType* outputMatrix,
433 void checkIndex( OrdinalType rowIndex, OrdinalType colIndex = 0 )
const;
434 OrdinalType numRows_;
435 OrdinalType numCols_;
448 template<
typename OrdinalType,
typename ScalarType>
451 BLAS<OrdinalType,ScalarType>(),
457 valuesCopied_ (false),
461 template<
typename OrdinalType,
typename ScalarType>
464 OrdinalType numCols_in,
469 BLAS<OrdinalType,ScalarType>(),
470 numRows_ (numRows_in),
471 numCols_ (numCols_in),
472 stride_ (kl_in+ku_in+1),
475 valuesCopied_ (true),
478 values_ =
new ScalarType[stride_ * numCols_];
484 template<
typename OrdinalType,
typename ScalarType>
487 ScalarType* values_in,
488 OrdinalType stride_in,
489 OrdinalType numRows_in,
490 OrdinalType numCols_in,
494 BLAS<OrdinalType,ScalarType>(),
495 numRows_ (numRows_in),
496 numCols_ (numCols_in),
500 valuesCopied_ (false),
505 values_ =
new ScalarType[stride_*numCols_];
506 copyMat (values_in, stride_in, numRows_, numCols_, values_, stride_, 0);
507 valuesCopied_ =
true;
511 template<
typename OrdinalType,
typename ScalarType>
515 BLAS<OrdinalType,ScalarType>(),
521 valuesCopied_ (true),
525 numRows_ = Source.numRows_;
526 numCols_ = Source.numCols_;
530 values_ =
new ScalarType[stride_*numCols_];
531 copyMat (Source.values_, Source.stride_, numRows_, numCols_,
532 values_, stride_, 0);
535 numRows_ = Source.numCols_;
536 numCols_ = Source.numRows_;
540 values_ =
new ScalarType[stride_*numCols_];
541 for (OrdinalType j = 0; j < numCols_; ++j) {
542 for (OrdinalType i = TEUCHOS_MAX(0,j-ku_);
543 i <= TEUCHOS_MIN(numRows_-1,j+kl_); ++i) {
544 values_[j*stride_ + (ku_+i-j)] =
550 numRows_ = Source.numCols_;
551 numCols_ = Source.numRows_;
555 values_ =
new ScalarType[stride_*numCols_];
556 for (OrdinalType j=0; j<numCols_; j++) {
557 for (OrdinalType i = TEUCHOS_MAX(0,j-ku_);
558 i <= TEUCHOS_MIN(numRows_-1,j+kl_); ++i) {
559 values_[j*stride_ + (ku_+i-j)] = Source.values_[i*Source.stride_ + (Source.ku_+j-i)];
565 template<
typename OrdinalType,
typename ScalarType>
569 OrdinalType numRows_in,
570 OrdinalType numCols_in,
571 OrdinalType startCol)
573 BLAS<OrdinalType,ScalarType>(),
574 numRows_ (numRows_in),
575 numCols_ (numCols_in),
576 stride_ (Source.stride_),
579 valuesCopied_ (false),
580 values_ (Source.values_)
583 values_ =
new ScalarType[stride_ * numCols_in];
584 copyMat (Source.values_, Source.stride_, numRows_in, numCols_in,
585 values_, stride_, startCol);
586 valuesCopied_ =
true;
588 values_ = values_ + (stride_ * startCol);
592 template<
typename OrdinalType,
typename ScalarType>
602 template<
typename OrdinalType,
typename ScalarType>
604 OrdinalType numRows_in, OrdinalType numCols_in, OrdinalType kl_in, OrdinalType ku_in
608 numRows_ = numRows_in;
609 numCols_ = numCols_in;
613 values_ =
new ScalarType[stride_*numCols_];
615 valuesCopied_ =
true;
620 template<
typename OrdinalType,
typename ScalarType>
622 OrdinalType numRows_in, OrdinalType numCols_in, OrdinalType kl_in, OrdinalType ku_in
626 numRows_ = numRows_in;
627 numCols_ = numCols_in;
631 values_ =
new ScalarType[stride_*numCols_];
632 valuesCopied_ =
true;
637 template<
typename OrdinalType,
typename ScalarType>
639 OrdinalType numRows_in, OrdinalType numCols_in, OrdinalType kl_in, OrdinalType ku_in
644 ScalarType* values_tmp =
new ScalarType[(kl_in+ku_in+1) * numCols_in];
646 for(OrdinalType k = 0; k < (kl_in+ku_in+1) * numCols_in; k++) {
647 values_tmp[k] = zero;
649 OrdinalType numRows_tmp = TEUCHOS_MIN(numRows_, numRows_in);
650 OrdinalType numCols_tmp = TEUCHOS_MIN(numCols_, numCols_in);
652 copyMat(values_, stride_, numRows_tmp, numCols_tmp, values_tmp,
656 numRows_ = numRows_in;
657 numCols_ = numCols_in;
661 values_ = values_tmp;
662 valuesCopied_ =
true;
671 template<
typename OrdinalType,
typename ScalarType>
676 for(OrdinalType j = 0; j < numCols_; j++) {
677 for (OrdinalType i=TEUCHOS_MAX(0,j-ku_); i<=TEUCHOS_MIN(numRows_-1,j+kl_); i++) {
678 values_[(ku_+i-j) + j*stride_] = value_in;
685 template<
typename OrdinalType,
typename ScalarType>
690 for(OrdinalType j = 0; j < numCols_; j++) {
691 for (OrdinalType i=TEUCHOS_MAX(0,j-ku_); i<=TEUCHOS_MIN(numRows_-1,j+kl_); i++) {
699 template<
typename OrdinalType,
typename ScalarType>
708 if((!valuesCopied_) && (!Source.valuesCopied_) && (values_ == Source.values_))
712 if (!Source.valuesCopied_) {
717 numRows_ = Source.numRows_;
718 numCols_ = Source.numCols_;
721 stride_ = Source.stride_;
722 values_ = Source.values_;
726 numRows_ = Source.numRows_;
727 numCols_ = Source.numCols_;
731 const OrdinalType newsize = stride_ * numCols_;
733 values_ =
new ScalarType[newsize];
734 valuesCopied_ =
true;
740 if((Source.numRows_ <= stride_) && (Source.numCols_ == numCols_)) {
741 numRows_ = Source.numRows_;
742 numCols_ = Source.numCols_;
748 numRows_ = Source.numRows_;
749 numCols_ = Source.numCols_;
753 const OrdinalType newsize = stride_ * numCols_;
755 values_ =
new ScalarType[newsize];
756 valuesCopied_ =
true;
760 copyMat(Source.values_, Source.stride_, numRows_, numCols_, values_, stride_, 0);
766 template<
typename OrdinalType,
typename ScalarType>
771 if ((numRows_ != Source.numRows_) || (numCols_ != Source.numCols_) || (kl_ != Source.kl_) || (ku_ != Source.ku_)) {
772 TEUCHOS_CHK_REF(*
this);
779 template<
typename OrdinalType,
typename ScalarType>
784 if ((numRows_ != Source.numRows_) || (numCols_ != Source.numCols_) || (kl_ != Source.kl_) || (ku_ != Source.ku_)) {
785 TEUCHOS_CHK_REF(*
this);
792 template<
typename OrdinalType,
typename ScalarType>
797 if((!valuesCopied_) && (!Source.valuesCopied_) && (values_ == Source.values_))
801 if ((numRows_ != Source.numRows_) || (numCols_ != Source.numCols_) || (kl_ != Source.kl_) || (ku_ != Source.ku_)) {
802 TEUCHOS_CHK_REF(*
this);
804 copyMat(Source.values_, Source.stride_, numRows_, numCols_, values_, stride_, 0);
813 template<
typename OrdinalType,
typename ScalarType>
816 #ifdef HAVE_TEUCHOS_ARRAY_BOUNDSCHECK
817 checkIndex( rowIndex, colIndex );
819 return(values_[colIndex * stride_ + ku_+rowIndex-colIndex]);
822 template<
typename OrdinalType,
typename ScalarType>
825 #ifdef HAVE_TEUCHOS_ARRAY_BOUNDSCHECK
826 checkIndex( rowIndex, colIndex );
828 return(values_[colIndex * stride_ + ku_+rowIndex-colIndex]);
831 template<
typename OrdinalType,
typename ScalarType>
834 #ifdef HAVE_TEUCHOS_ARRAY_BOUNDSCHECK
835 checkIndex( 0, colIndex );
837 return(values_ + colIndex * stride_);
840 template<
typename OrdinalType,
typename ScalarType>
843 #ifdef HAVE_TEUCHOS_ARRAY_BOUNDSCHECK
844 checkIndex( 0, colIndex );
846 return(values_ + colIndex * stride_);
853 template<
typename OrdinalType,
typename ScalarType>
861 for(j = 0; j < numCols_; j++) {
863 ptr = values_ + j * stride_ + TEUCHOS_MAX(0, ku_-j);
864 for (i=TEUCHOS_MAX(0,j-ku_); i<=TEUCHOS_MIN(numRows_-1,j+kl_); i++) {
872 updateFlops((kl_+ku_+1) * numCols_);
877 template<
typename OrdinalType,
typename ScalarType>
883 for (i = 0; i < numRows_; i++) {
885 for (j=TEUCHOS_MAX(0,i-kl_); j<=TEUCHOS_MIN(numCols_-1,i+ku_); j++) {
888 anorm = TEUCHOS_MAX( anorm, sum );
890 updateFlops((kl_+ku_+1) * numCols_);
895 template<
typename OrdinalType,
typename ScalarType>
901 for (j = 0; j < numCols_; j++) {
902 for (i=TEUCHOS_MAX(0,j-ku_); i<=TEUCHOS_MIN(numRows_-1,j+kl_); i++) {
907 updateFlops((kl_+ku_+1) * numCols_);
916 template<
typename OrdinalType,
typename ScalarType>
921 if((numRows_ != Operand.numRows_) || (numCols_ != Operand.numCols_) || (kl_ != Operand.kl_) || (ku_ != Operand.ku_)) {
925 for(j = 0; j < numCols_; j++) {
926 for (i=TEUCHOS_MAX(0,j-ku_); i<=TEUCHOS_MIN(numRows_-1,j+kl_); i++) {
927 if((*
this)(i, j) != Operand(i, j)) {
937 template<
typename OrdinalType,
typename ScalarType>
940 return !((*this) == Operand);
947 template<
typename OrdinalType,
typename ScalarType>
950 this->scale( alpha );
954 template<
typename OrdinalType,
typename ScalarType>
961 for (j=0; j<numCols_; j++) {
962 ptr = values_ + j*stride_ + TEUCHOS_MAX(0, ku_-j);
963 for (i=TEUCHOS_MAX(0,j-ku_); i<=TEUCHOS_MIN(numRows_-1,j+kl_); i++) {
964 *ptr = alpha * (*ptr); ptr++;
967 updateFlops( (kl_+ku_+1)*numCols_ );
972 template<
typename OrdinalType,
typename ScalarType>
980 if ((numRows_ != A.numRows_) || (numCols_ != A.numCols_) || (kl_ != A.kl_) || (ku_ != A.ku_)) {
983 for (j=0; j<numCols_; j++) {
984 ptr = values_ + j*stride_ + TEUCHOS_MAX(0, ku_-j);
985 for (i=TEUCHOS_MAX(0,j-ku_); i<=TEUCHOS_MIN(numRows_-1,j+kl_); i++) {
986 *ptr = A(i,j) * (*ptr); ptr++;
989 updateFlops( (kl_+ku_+1)*numCols_ );
994 template<
typename OrdinalType,
typename ScalarType>
999 os <<
"Values_copied : yes" << std::endl;
1001 os <<
"Values_copied : no" << std::endl;
1002 os <<
"Rows : " << numRows_ << std::endl;
1003 os <<
"Columns : " << numCols_ << std::endl;
1004 os <<
"Lower Bandwidth : " << kl_ << std::endl;
1005 os <<
"Upper Bandwidth : " << ku_ << std::endl;
1006 os <<
"LDA : " << stride_ << std::endl;
1007 if(numRows_ == 0 || numCols_ == 0) {
1008 os <<
"(matrix is empty, no values to display)" << std::endl;
1011 for(OrdinalType i = 0; i < numRows_; i++) {
1012 for (OrdinalType j=TEUCHOS_MAX(0,i-kl_); j<=TEUCHOS_MIN(numCols_-1,i+ku_); j++) {
1013 os << (*this)(i,j) <<
" ";
1024 template<
typename OrdinalType,
typename ScalarType>
1028 rowIndex < TEUCHOS_MAX(0,colIndex-ku_) || rowIndex > TEUCHOS_MIN(numRows_-1,colIndex+kl_),
1030 "SerialBandDenseMatrix<T>::checkIndex: "
1031 "Row index " << rowIndex <<
" out of range [0, "<< numRows_ <<
")");
1033 "SerialBandDenseMatrix<T>::checkIndex: "
1034 "Col index " << colIndex <<
" out of range [0, "<< numCols_ <<
")");
1038 template<
typename OrdinalType,
typename ScalarType>
1039 void SerialBandDenseMatrix<OrdinalType, ScalarType>::deleteArrays(
void)
1041 if (valuesCopied_) {
1044 valuesCopied_ =
false;
1048 template<
typename OrdinalType,
typename ScalarType>
1049 void SerialBandDenseMatrix<OrdinalType, ScalarType>::copyMat(
1050 ScalarType* inputMatrix, OrdinalType strideInput, OrdinalType numRows_in,
1051 OrdinalType numCols_in, ScalarType* outputMatrix, OrdinalType strideOutput, OrdinalType startCol, ScalarType alpha
1055 ScalarType* ptr1 = 0;
1056 ScalarType* ptr2 = 0;
1058 for(j = 0; j < numCols_in; j++) {
1059 ptr1 = outputMatrix + (j * strideOutput) + TEUCHOS_MAX(0, ku_-j);
1060 ptr2 = inputMatrix + (j + startCol) * strideInput + TEUCHOS_MAX(0, ku_-j);
1062 for (i=TEUCHOS_MAX(0,j-ku_); i<=TEUCHOS_MIN(numRows_in-1,j+kl_); i++) {
1063 *ptr1++ += alpha*(*ptr2++);
1066 for (i=TEUCHOS_MAX(0,j-ku_); i<=TEUCHOS_MIN(numRows_in-1,j+kl_); i++) {
1073 #ifndef TEUCHOS_HIDE_DEPRECATED_CODE
1074 template<
typename OrdinalType,
typename ScalarType>
1076 std::ostream& operator<< (std::ostream& os, const Teuchos::SerialBandDenseMatrix<OrdinalType, ScalarType>& obj)
bool empty() const
Returns whether this matrix is empty.
OrdinalType numRows() const
Returns the row dimension of this matrix.
SerialBandDenseMatrix< OrdinalType, ScalarType > & operator+=(const SerialBandDenseMatrix< OrdinalType, ScalarType > &Source)
Add another matrix to this matrix.
ScalarType & operator()(OrdinalType rowIndex, OrdinalType colIndex)
Element access method (non-const).
int scale(const ScalarType alpha)
Scale this matrix by alpha; *this = alpha**this.
int shape(OrdinalType numRows, OrdinalType numCols, OrdinalType kl, OrdinalType ku)
Shape method for changing the size of a SerialBandDenseMatrix, initializing entries to zero...
ScalarTraits< ScalarType >::magnitudeType normInf() const
Returns the Infinity-norm of the matrix.
Templated interface class to BLAS routines.
virtual ~SerialBandDenseMatrix()
Destructor.
#define TEUCHOS_TEST_FOR_EXCEPTION(throw_exception_test, Exception, msg)
Macro for throwing an exception with breakpointing to ease debugging.
int reshape(OrdinalType numRows, OrdinalType numCols, OrdinalType kl, OrdinalType ku)
Reshaping method for changing the size of a SerialBandDenseMatrix, keeping the entries.
Teuchos header file which uses auto-configuration information to include necessary C++ headers...
ScalarTraits< ScalarType >::magnitudeType normOne() const
Returns the 1-norm of the matrix.
SerialBandDenseMatrix< OrdinalType, ScalarType > & assign(const SerialBandDenseMatrix< OrdinalType, ScalarType > &Source)
Copies values from one matrix to another.
Object for storing data and providing functionality that is common to all computational classes...
This structure defines some basic traits for a scalar field type.
OrdinalType stride() const
Returns the stride between the columns of this matrix in memory.
virtual void print(std::ostream &os) const
Print method. Defines the behavior of the std::ostream << operator.
ScalarType * values() const
Data array access method.
Functionality and data that is common to all computational classes.
bool operator!=(const SerialBandDenseMatrix< OrdinalType, ScalarType > &Operand) const
Inequality of two matrices.
ScalarTraits< ScalarType >::magnitudeType normFrobenius() const
Returns the Frobenius-norm of the matrix.
This class creates and provides basic support for banded dense matrices of templated type...
SerialBandDenseMatrix< OrdinalType, ScalarType > & operator*=(const ScalarType alpha)
Scale this matrix by alpha; *this = alpha**this.
SerialBandDenseMatrix< OrdinalType, ScalarType > & operator-=(const SerialBandDenseMatrix< OrdinalType, ScalarType > &Source)
Subtract another matrix from this matrix.
int putScalar(const ScalarType value=Teuchos::ScalarTraits< ScalarType >::zero())
Set all values in the matrix to a constant value.
int random()
Set all values in the matrix to be random numbers.
SerialBandDenseMatrix()
Default Constructor.
bool operator==(const SerialBandDenseMatrix< OrdinalType, ScalarType > &Operand) const
Equality of two matrices.
static magnitudeType magnitude(T a)
Returns the magnitudeType of the scalar type a.
OrdinalType ordinalType
Typedef for ordinal type.
SerialBandDenseMatrix< OrdinalType, ScalarType > & operator=(const SerialBandDenseMatrix< OrdinalType, ScalarType > &Source)
Copies values from one matrix to another.
OrdinalType upperBandwidth() const
Returns the upper bandwidth of this matrix.
Defines basic traits for the scalar field type.
static T zero()
Returns representation of zero for this scalar type.
OrdinalType lowerBandwidth() const
Returns the lower bandwidth of this matrix.
ScalarType * operator[](OrdinalType colIndex)
Column access method (non-const).
ScalarType scalarType
Typedef for scalar type.
OrdinalType numCols() const
Returns the column dimension of this matrix.
Reference-counted pointer class and non-member templated function implementations.
int shapeUninitialized(OrdinalType numRows, OrdinalType numCols, OrdinalType kl, OrdinalType ku)
Same as shape() except leaves uninitialized.
Teuchos::DataAccess Mode enumerable type.