LA_library/sparsesmat.h
2011-01-18 14:37:05 +00:00

607 lines
18 KiB
C++

/*
LA: linear algebra C++ interface library
Copyright (C) 2008 Jiri Pittner <jiri.pittner@jh-inst.cas.cz> or <jiri@pittnerovi.com>
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#ifndef _SPARSESMAT_H_
#define _SPARSESMAT_H_
#include <string>
#include <cmath>
#include <stdlib.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <errno.h>
#include "la_traits.h"
#include "sparsemat.h"
#include "vec.h"
#include "mat.h"
#include "smat.h"
#include "qsort.h"
#include <map>
#include <list>
#define CHOLESKYEPSILON 1e-16
namespace LA {
//symmetric sparse matrix class with a representation for efficient exponentiatiation
//in particular we need a unitary symmetric complex matrix as exp(iH) with H real symmetric
//indices are counted from zero
template <typename T>
class SparseSMat
{
protected:
SPMatindex nn;
SPMatindex mm;
std::map<SPMatindex,T> **v;
int *count;
public:
SparseSMat() : nn(0), mm(0), v(NULL), count(NULL) {};
explicit SparseSMat(const SPMatindex n, const SPMatindex m); //prevent double -> int -> SparseSMat
explicit SparseSMat(const SPMatindex n);
SparseSMat(const SparseSMat &rhs);
explicit SparseSMat(const SparseMat<T> &rhs);
explicit SparseSMat(const NRSMat<T> &rhs);
explicit SparseSMat(const NRMat<T> &rhs);
explicit SparseSMat(const CSRMat<T> &rhs);
SparseSMat & operator=(const SparseSMat &rhs);
void copyonwrite();
void resize(const SPMatindex nn, const SPMatindex mm);
void dealloc(void) {resize(0,0);}
inline void setcoldim(int i) {mm=(SPMatindex)i;};
//std::map<SPMatindex,T> *line(SPMatindex n) const {return v[n];};
typedef std::map<SPMatindex,T> *ROWTYPE;
inline typename SparseSMat<T>::ROWTYPE & operator[](const SPMatindex i) {return v[i];};
void clear() {resize(nn,mm);}
unsigned long long simplify();
~SparseSMat();
inline int getcount() const {return count?*count:0;}
SparseSMat & operator*=(const T &a); //multiply by a scalar
inline const SparseSMat operator*(const T &rhs) const {return SparseSMat(*this) *= rhs;}
inline const SparseSMat operator+(const T &rhs) const {return SparseSMat(*this) += rhs;}
inline const SparseSMat operator-(const T &rhs) const {return SparseSMat(*this) -= rhs;}
inline const SparseSMat operator+(const SparseSMat &rhs) const {return SparseSMat(*this) += rhs;}
inline const SparseSMat operator-(const SparseSMat &rhs) const {return SparseSMat(*this) -= rhs;}
SparseSMat & operator=(const T &a); //assign a to diagonal
SparseSMat & operator+=(const T &a); //assign a to diagonal
SparseSMat & operator-=(const T &a); //assign a to diagonal
void axpy(const T alpha, const SparseSMat &x, const bool transp=0); // this+= a*x
inline SparseSMat & operator+=(const SparseSMat &rhs) {axpy((T)1,rhs); return *this;};
inline SparseSMat & operator-=(const SparseSMat &rhs) {axpy((T)-1,rhs); return *this;};
const T* diagonalof(NRVec<T> &, const bool divide=0, bool cache=false) const; //get diagonal
void gemv(const T beta, NRVec<T> &r, const char trans, const T alpha, const NRVec<T> &x) const;
inline const NRVec<T> operator*(const NRVec<T> &rhs) const {NRVec<T> result(nn); this->gemv((T)0,result,'n',(T)1,rhs); return result;};
typename LA_traits<T>::normtype norm(const T scalar=(T)0) const;
inline const SparseSMat operator*(const SparseSMat &rhs) const {SparseSMat<T> r(nn,mm); r.gemm(0,*this,'n',rhs,'n',1); return r;}; //!!!NOT A GENERAL ROUTINE, JUST FOR THE CASES WHEN THE RESULT STAYS SYMMETRIC
void gemm(const T beta, const SparseSMat &a, const char transa, const SparseSMat &b, const char transb, const T alpha); //this := alpha*op( A )*op( B ) + beta*this !!!NOT A GENERAL ROUTINE, JUST FOR THE CASES WHEN THE RESULT STAYS SYMMETRIC
inline void add(const SPMatindex n, const SPMatindex m, const T elem, const bool both=true);
inline unsigned long long length() {return simplify();};
void transposeme() const {laerror("in-place transposition not necessary/implemented for SparseSMat");};
SparseSMat transpose(bool conj=false) const; //if we store a non-symmetric matrix there
inline bool issymmetric() const {return true;} // LV: for davidson
void get(int fd, bool dimen, bool transp);
void put(int fd, bool dimen, bool transp) const;
int nrows() const {return nn;}
int ncols() const {return mm;}
SparseSMat<T> cholesky(void) const;
SparseSMat submatrix(const int fromrow, const int torow, const int fromcol, const int tocol) const;
void storesubmatrix(const int fromrow, const int fromcol, const SparseSMat &rhs);
class iterator {//not efficient, just for output to ostreams
private:
matel<T> *p;
matel<T> my;
SPMatindex row;
typename std::map<SPMatindex,T>::iterator *col;
typename std::map<SPMatindex,T>::iterator mycol;
SPMatindex mynn;
SPMatindex mymm;
std::map<SPMatindex,T> **myv;
public:
//compiler-generated iterator & operator=(const iterator &rhs);
//compiler-generated iterator(const iterator &rhs);
iterator(): p(NULL),row(0),col(NULL),mynn(0),mymm(0),myv(NULL) {};
iterator(const SparseSMat &rhs) : mynn(rhs.nn), mymm(rhs.mm), myv(rhs.v), col(NULL) {row=0; p= &my; operator++();}
iterator operator++() {
if(col) //finish column list
{
++mycol;
if(mycol != myv[row]->end())
{
p->row = row;
p->col = mycol->first;
p->elem = mycol->second;
return *this;
}
else
{
col=NULL;
++row; if(row==mynn) {p=NULL; return *this;} //end()
}
}
nextrow:
while(myv[row]==NULL) {++row; if(row==mynn) {p=NULL; return *this;}} //end()
//we are at next nonempty row
col = &mycol;
mycol = myv[row]->begin();
if(mycol == myv[row]->end()) {col=NULL;
++row;
if(row==mynn) {p=NULL; return *this;} else goto nextrow;
}
//first column of new row
p->row = row;
p->col = mycol->first;
p->elem = mycol->second;
return *this;
};
iterator(matel<T> *q) {p=q; col=NULL;}//just for end()
//compiler-generated ~iterator() {};
bool operator!=(const iterator &rhs) const {return p!=rhs.p;} //only for comparison with end()
bool operator==(const iterator &rhs) const {return p==rhs.p;} //only for comparison with end()
matel<T> & operator*() const {return *p;}
matel<T> * operator->() const {return p;}
bool notend() const {return (bool)p;};
};
iterator begin() const {return iterator(*this);}
iterator end() const {return iterator(NULL);}
};
template <typename T>
SparseSMat<T>::SparseSMat(const SPMatindex n)
:nn(n), mm(n),
count(new int(1))
{
v= new std::map<SPMatindex,T> * [n];
memset(v,0,nn*sizeof(std::map<SPMatindex,T> *));
}
template <typename T>
SparseSMat<T>::SparseSMat(const SPMatindex n, const SPMatindex m)
:nn(n), mm(m),
count(new int(1))
{
v= new std::map<SPMatindex,T> * [n];
memset(v,0,nn*sizeof(std::map<SPMatindex,T> *));
}
template <typename T>
SparseSMat<T>::SparseSMat(const NRSMat<T> &rhs)
:nn(rhs.nrows()), mm(rhs.ncols()),
count(new int(1))
{
v= new std::map<SPMatindex,T> * [nn];
memset(v,0,nn*sizeof(std::map<SPMatindex,T> *));
int i,j;
for(i=0; i<nn; ++i) for(j=0; j<=i; ++j) if(std::abs(rhs(i,j))>SPARSEEPSILON) (*this).add(i,j,rhs(i,j),true);
}
template <typename T>
SparseSMat<T>::SparseSMat(const NRMat<T> &rhs)
:nn(rhs.nrows()), mm(rhs.ncols()),
count(new int(1))
{
if(rhs.nrows()!=rhs.ncols()) laerror("non-square matrix in SparseSMat constructor from NRMat");
v= new std::map<SPMatindex,T> * [nn];
memset(v,0,nn*sizeof(std::map<SPMatindex,T> *));
int i,j;
for(i=0; i<nn; ++i) for(j=0; j<mm; ++j) if(std::abs(rhs(i,j))>SPARSEEPSILON) (*this).add(i,j,rhs(i,j),false);
}
template <typename T>
SparseSMat<T>::SparseSMat(const SparseSMat &rhs)
{
v = rhs.v;
nn = rhs.nn;
mm = rhs.mm;
count = rhs.count;
if(count) (*count)++;
}
//NRSMat from SparseSMat
#define nn2 (nn*(nn+1)/2)
template <typename T>
NRSMat<T>::NRSMat(const SparseSMat<T> &rhs)
: nn(rhs.nrows())
{
if(rhs.nrows()!=rhs.ncols()) laerror("cannot transform rectangular matrix to NRSMat");
#ifdef CUDALA
location = cpu;
#endif
count = new int(1);
v=new T[nn2];
memset(v,0,nn2*sizeof(T));
typename SparseSMat<T>::iterator p(rhs);
for(; p.notend(); ++p) if(p->row <= p->col) (*this)(p->row,p->col)=p->elem;
}
#undef nn2
//construct dense from sparse
template <typename T>
NRMat<T>::NRMat(const SparseSMat<T> &rhs) :
nn(rhs.nrows()),
mm(rhs.ncols()),
count(new int(1))
{
#ifdef CUDALA
location = cpu;
#endif
#ifdef MATPTR
v = new T*[nn];
v[0] = new T[mm*nn];
for (int i=1; i<nn; i++) v[i] = v[i-1] + mm;
#else
v = new T[mm*nn];
#endif
memset(&(*this)(0,0),0,mm*nn*sizeof(T));
typename SparseSMat<T>::iterator p(rhs);
for(; p.notend(); ++p) (*this)(p->row,p->col)= p->elem;
}
template <typename T>
SparseSMat<T>::~SparseSMat()
{
if(!count) return;
if(--(*count) <= 0) {
if(v)
{
for(SPMatindex i=0; i<nn; ++i) if(v[i]) delete v[i];
delete[] (v);
}
delete count;
}
}
template <typename T>
void SparseSMat<T>::resize(const SPMatindex n, const SPMatindex m)
{
if(!count)
{
if(n==0) return;
count = new int(1);
nn=n;
mm=m;
v= new std::map<SPMatindex,T> * [nn];
for(SPMatindex i=0; i<nn; ++i) v[i]=NULL;
return;
}
if(*count > 1) //it was shared
{
(*count)--;
if(n)
{
count = new int(1);
nn=n;
mm=m;
v= new std::map<SPMatindex,T> * [nn];
for(SPMatindex i=0; i<nn; ++i) v[i]=NULL;
}
else {v=NULL; nn=0; mm=0; count=NULL;}
}
else //it was not shared
{
mm=m;
//delete all trees
for(SPMatindex i=0; i<nn; ++i) if(v[i]) {delete v[i]; v[i]=NULL;}
if(n!=nn)
{
nn=n;
for(SPMatindex i=0; i<nn; ++i) v[i]=NULL;
}
}
}
template <typename T>
SparseSMat<T> & SparseSMat<T>::operator=(const SparseSMat &rhs)
{
if (this != &rhs)
{
if(count)
if(--(*count) == 0)
{
if(v)
{
for(SPMatindex i=0; i<nn; ++i) if(v[i]) delete v[i];
delete[] (v);
}
delete count;
}
v = rhs.v;
nn = rhs.nn;
mm = rhs.mm;
count = rhs.count;
if(count) (*count)++;
}
return *this;
}
template <typename T>
void SparseSMat<T>::copyonwrite()
{
if(!count) laerror("SparseSmat::copyonwrite() of an undefined object");
if(*count > 1)
{
(*count)--;
count = new int;
*count = 1;
typename std::map<SPMatindex,T> **newv= new std::map<SPMatindex,T> * [nn];
for(SPMatindex i=0; i<nn; ++i) if(v[i])
newv[i]= new typename std::map<SPMatindex,T>(*(v[i])); //deep copy of each map
else
newv[i]= NULL;
v = newv;
}
}
template <typename T>
void SparseSMat<T>::add(const SPMatindex n, const SPMatindex m, const T elem, const bool both)
{
#ifdef DEBUG
if(n>=nn || m>=mm) laerror("illegal index in SparseSMat::add()");
#endif
if(!v[n]) v[n] = new std::map<SPMatindex,T>;
typename std::map<SPMatindex,T>::iterator p;
p= v[n]->find(m);
if(p!=v[n]->end()) p->second+=elem; else (*v[n])[m] = elem;
if(n!=m && both) //add also transposed
{
if(!v[m]) v[m] = new std::map<SPMatindex,T>;
p= v[m]->find(n);
if(p!=v[m]->end()) p->second+=elem; else (*v[m])[n] = elem;
}
//addition can lead to zero, but do not treat it now, make a simplify
}
template <typename T>
unsigned long long SparseSMat<T>::simplify()
{
unsigned long long count=0;
for(SPMatindex i=0; i<nn; ++i) if(v[i])
{
//check for zero elements and erase them from the list
//build a list since we are not sure whether erase from within the traversal loop is safe
std::list<SPMatindex> l;
typename std::map<SPMatindex,T>::iterator p;
for(p=v[i]->begin(); p!=v[i]->end(); ++p)
if(std::abs(p->second) < SPARSEEPSILON) l.push_front(p->first); else ++count;
typename std::list<SPMatindex>::iterator q;
for(q=l.begin(); q!=l.end(); ++q) v[i]->erase(*q);
if(v[i]->size() == 0) {delete v[i]; v[i]=NULL;}
}
return count;
}
template <typename T>
std::ostream & operator<<(std::ostream &s, const SparseSMat<T> &x)
{
SPMatindex n;
s << x.nrows() << " "<< x.ncols()<< std::endl;
typename SparseSMat<T>::iterator p(x);
for(; p.notend(); ++p) s << (int)p->row << ' ' << (int)p->col << ' ' << (typename LA_traits_io<T>::IOtype) p->elem << '\n';
s << "-1 -1\n";
return s;
}
template <class T>
std::istream& operator>>(std::istream &s, SparseSMat<T> &x)
{
SPMatindex n,m;
long i,j;
s >> n >> m;
if(n!=m) laerror("SparseSMat must be square");
x.resize(n,m);
s >> i >> j;
typename LA_traits_io<T>::IOtype tmp;
while(i>=0 && j>=0)
{
s>>tmp;
if(i>=n||j>=m) laerror("bad index in SparseSMat::operator>>");
x.add(i,j,tmp,false);
s >> i >> j;
}
return s;
}
template <typename T>
SparseSMat<T> SparseSMat<T>::transpose(bool conj) const
{
SparseSMat<T> r(mm,nn);
typename SparseSMat<T>::iterator p(*this);
for(; p.notend(); ++p) r.add(p->col, p->row, (conj?LA_traits<T>::conjugate(p->elem):p->elem), false);
return r;
}
//Cholesky decomposition, pivoted, positive semidefinite, not in place
//it is NOT checked that the input matrix is symmetric/hermitean
//result.transpose(true)*result reproduces the original matrix
//Due to pivoting the result is upper triangular only before applying final permutation
//
template <typename T>
SparseSMat<T> SparseSMat<T>::cholesky(void) const
{
if(nn!=mm) laerror("Cholesky defined only for square matrices");
//we need real values for sorting, if T is already real it makes just an unnecessary copy of one vector
NRVec<typename LA_traits<T>::normtype> diagreal(nn);
{
NRVec<T> diag(nn); diagonalof(diag);
for(SPMatindex i=0; i<nn; ++i) diagreal[i]=LA_traits<T>::realpart(diag[i]);
}
NRVec<int> pivot(nn);
for(int i=0; i<nn; ++i) pivot[i]=i;
//pivot by sorting
//!this is actually not fully correct approach, since the pivoting should be done during the Cholesky process
//Now it can happen that some elements will vanish in the process, while there will be some remaining ones later
//However, column swapping in the regular pivoting in an in-place algorithm would be rather clumsy with std::map , since simply renumbering the key is not allowed
//This works reasonably well so keep it like this at the moment
diagreal.sort(1,0,nn-1,pivot);
//prepare inverse permutation
NRVec<int> invpivot(nn);
for(int i=0; i<nn; ++i) invpivot[pivot[i]]=i;
//std::cout <<"sorted diagonal\n"<<diagreal;
//std::cout <<"pivot\n"<<pivot;
//copy-permute upper triangle
SparseSMat<T> r;
r.nn=nn;
r.mm=nn;
r.count = new int(1);
r.v = new std::map<SPMatindex,T> * [nn];
for(SPMatindex i=0; i<nn; ++i)
if(v[pivot[i]])
{
r.v[i]= new typename std::map<SPMatindex,T>;
typename std::map<SPMatindex,T>::iterator p;
for(p=v[pivot[i]]->begin(); p!=v[pivot[i]]->end(); ++p)
{
if(invpivot[p->first] >= i)
{
(*r.v[i])[invpivot[p->first]] = p->second;
}
}
}
else
r.v[i]= NULL;
//std::cout <<"Permuted upper triangle matrix\n"<<r;
//SparseSMat<T> r0(r);r.copyonwrite();
//perform complex, positive semidefinite Cholesky with thresholding by SPARSEEPSILON
SPMatindex i,j,k;
int rank=0;
for(k=0; k<nn; ++k)
if(r.v[k])
{
typename std::map<SPMatindex,T>::iterator p;
p= r.v[k]->find(k);
if(p==r.v[k]->end()) continue; //must not break due to the a priori pivoting
if(LA_traits<T>::realpart(p->second) < CHOLESKYEPSILON) continue; //must not break due to the a priori pivoting
++rank;
typename LA_traits<T>::normtype tmp = std::sqrt(LA_traits<T>::realpart(p->second));
p->second = tmp;
NRVec<T> linek(0.,nn);
for(p=r.v[k]->begin(); p!=r.v[k]->end(); ++p)
{
if(p->first > k) p->second /= tmp;
linek[p->first] = p->second;
}
for(j=k+1; j<nn; ++j)
if(r.v[j])
{
T akj = LA_traits<T>::conjugate(linek[j]);
NRVec<int> linek_used(0,nn);
if(std::abs(akj)>SPARSEEPSILON)
{
for(p=r.v[j]->begin(); p!=r.v[j]->end(); ++p)
{
p->second -= akj * linek[p->first];
linek_used[p->first]=1;
}
//subtract also elements nonzero in line k but non-existent in line j
for(i=j; i<nn; ++i)
if(!linek_used[i] && std::abs(linek[i]) > SPARSEEPSILON)
{
(*r.v[j])[i] = -akj * linek[i];
}
}
}
}
/*
SparseSMat<T> br(nn);
br.gemm(0,r,'c',r,'n',1);
//cancel low triangle from br
for(k=0; k<nn; ++k)
if(br.v[k])
{
typename std::map<SPMatindex,T>::iterator p;
for(p=br.v[k]->begin(); p!=br.v[k]->end(); ++p)
if(p->first <k) p->second=0.;
}
std::cout << "Error before permute = " <<(br-r0).norm()<<std::endl;
*/
//permute the result back;
for(k=0; k<nn; ++k)
if(r.v[k])
{
typename std::map<SPMatindex,T>::iterator p;
typename std::map<SPMatindex,T> *vnew = new typename std::map<SPMatindex,T>;
for(p=r.v[k]->begin(); p!=r.v[k]->end(); ++p)
{
if(std::abs(p->second) > SPARSEEPSILON) (*vnew)[pivot[p->first]] = p->second;
}
delete r.v[k];
r.v[k]=vnew;
}
return r;
}
//outer product expected to be sparse
template<typename T>
SparseSMat<T> otimes_sparse(const NRVec<T> &lhs, const NRVec<T> &rhs, const bool conjugate=false, const T &scale=1)
{
SparseSMat<T> r(lhs.size(),rhs.size());
for(SPMatindex i=0; i<lhs.size(); ++i)
if(lhs[i]!=(T)0)
{
for(SPMatindex j=0; j<rhs.size(); ++j)
if(rhs[j]!=(T)0)
{
T x=lhs[i]*(conjugate?LA_traits<T>::conjugate(rhs[j]):rhs[j])*scale;
if(std::abs(x)>SPARSEEPSILON) r.add(i,j,x);
}
}
return r;
}
}//namespace
#endif //_SPARSESMAT_H_