Skip to content

GitLab

Commit dea8257e authored by Martin Marinov's avatar Martin Marinov
Browse files

Merge remote-tracking branch 'origin/warning_fixes'

parents 957aeb32 54dc54db
Hide whitespace changes
Inline Side-by-side
Solver/CholmodSolver.cc
View file @ dea8257e
......@@ -82,7 +82,7 @@ bool CholmodSolver::calc_system( const std::vector<int>& _colptr,
rowind_ = _rowind;
values_ = _values;
int n = colptr_.size()-1;
size_t n = colptr_.size()-1;
cholmod_sparse matA;
......@@ -167,7 +167,7 @@ bool CholmodSolver::calc_system_prepare_pattern( const std::vector<int>& _col
rowind_ = _rowind;
values_ = _values;
int n = colptr_.size()-1;
size_t n = colptr_.size() - 1;
// setup matrix matA
cholmod_sparse matA;
......@@ -282,7 +282,7 @@ bool CholmodSolver::update_system( const std::vector<int>& _colptr,
colptr_ = _colptr;
rowind_ = _rowind;
values_ = _values;
int n = colptr_.size()-1;
size_t n = colptr_.size() - 1;
cholmod_sparse matA;
......@@ -329,7 +329,7 @@ bool CholmodSolver::update_downdate_factor( const std::vector<int>& _colptr,
rowind_ = _rowind;
values_ = _values;
int n = colptr_.size()-1;
size_t n = colptr_.size() - 1;
cholmod_sparse matA;
......@@ -387,7 +387,7 @@ bool CholmodSolver::update_downdate_factor( const std::vector<int>& _colptr,
bool CholmodSolver::solve( double * _x, double * _b)
{
const unsigned int n = mp_L->n;
const size_t n = mp_L->n;
cholmod_dense *x, b;
......
Solver/ConstrainedSolver.hh
View file @ dea8257e
......@@ -303,13 +303,13 @@ public:
private:
template<class RowT, class MatrixT>
void add_row( int _row_i,
double _coeff,
RowT _row,
MatrixT& _mat );
void add_row( gmm::size_type _row_i,
double _coeff,
RowT _row,
MatrixT& _mat );
template<class RowT, class RMatrixT, class CMatrixT>
void add_row_simultaneously( int _row_i,
void add_row_simultaneously( gmm::size_type _row_i,
double _coeff,
RowT _row,
RMatrixT& _rmat,
......
Solver/ConstrainedSolverT.cc
View file @ dea8257e
......@@ -217,10 +217,9 @@ solve(
if( _show_miso_settings)
miso_.show_options_dialog();
int nrows = gmm::mat_nrows(_A);
int ncols = gmm::mat_ncols(_A);
int ncons = gmm::mat_nrows(_constraints);
gmm::size_type nrows = gmm::mat_nrows(_A);
gmm::size_type ncols = gmm::mat_ncols(_A);
gmm::size_type ncons = gmm::mat_nrows(_constraints);
DEB_out_if( _show_timings, 1, "Initital dimension: " << nrows << " x " << ncols
<< ", number of constraints: " << ncons
......@@ -290,8 +289,8 @@ resolve(
// COMISO_GMM::factored_to_quadratic(_B, A, rhs);
//TODO only compute rhs, not complete A for efficiency
unsigned int m = gmm::mat_nrows(_B);
unsigned int n = gmm::mat_ncols(_B);
gmm::size_type m = gmm::mat_nrows(_B);
gmm::size_type n = gmm::mat_ncols(_B);
typedef typename gmm::linalg_traits<RMatrixT>::const_sub_row_type CRowT;
typedef typename gmm::linalg_traits<RMatrixT>::sub_row_type RowT;
......@@ -349,8 +348,8 @@ resolve(
gmm::mult(rhs_update_table_.D_, *_constraint_rhs, rhs_update_table_.cur_constraint_rhs_);
// update rhs of stored constraints
unsigned int nc = gmm::mat_ncols(rhs_update_table_.constraints_p_);
for(unsigned int i=0; i<rhs_update_table_.cur_constraint_rhs_.size(); ++i)
gmm::size_type nc = gmm::mat_ncols(rhs_update_table_.constraints_p_);
for(gmm::size_type i=0; i<rhs_update_table_.cur_constraint_rhs_.size(); ++i)
rhs_update_table_.constraints_p_(i,nc-1) = -rhs_update_table_.cur_constraint_rhs_[i];
}
if(_rhs)
......@@ -397,14 +396,14 @@ make_constraints_independent(
{
DEB_enter_func;
// setup linear transformation for rhs, start with identity
unsigned int nr = gmm::mat_nrows(_constraints);
gmm::size_type nr = gmm::mat_nrows(_constraints);
gmm::resize(rhs_update_table_.D_, nr, nr);
gmm::clear(rhs_update_table_.D_);
for(unsigned int i=0; i<nr; ++i) rhs_update_table_.D_(i,i) = 1.0;
for(gmm::size_type i=0; i<nr; ++i) rhs_update_table_.D_(i,i) = 1.0;
// Base::StopWatch sw;
// number of variables
int n_vars = gmm::mat_ncols(_constraints);
const gmm::size_type n_vars = gmm::mat_ncols(_constraints);
// TODO Check: HZ added 14.08.09
_c_elim.clear();
......@@ -453,7 +452,7 @@ make_constraints_independent(
for(; row_it != row_end; ++row_it)
{
int cur_j = row_it.index();
int cur_j = static_cast<int>(row_it.index());
// do not use the constant part
if( cur_j != n_vars - 1 )
{
......@@ -462,7 +461,7 @@ make_constraints_independent(
{
if( fabs(*row_it) > max_elim_val)
{
elim_j = cur_j;
elim_j = (int)cur_j;
max_elim_val = fabs(*row_it);
}
//break;
......@@ -473,19 +472,18 @@ make_constraints_independent(
// gcd
// if the coefficient of an integer variable is not an integer, then
// the variable most problably will not be (expect if all coeffs are the same, e.g. 0.5)
if( (double(int(cur_row_val))- cur_row_val) != 0.0)
{
// std::cerr << __FUNCTION__ << " Warning: coefficient of integer variable is NOT integer: "
// << cur_row_val << std::endl;
gcd_update_valid = false;
}
if ((double(int(cur_row_val))- cur_row_val) != 0.0)
{
DEB_warning(2, "coefficient of integer variable is NOT integer : " << cur_row_val)
gcd_update_valid = false;
}
v_gcd[n_ints] = cur_row_val;
v_gcd[n_ints] = static_cast<int>(cur_row_val);
++n_ints;
// store integer closest to 1, must be greater than epsilon_
if( fabs(cur_row_val-1.0) < elim_val && cur_row_val > epsilon_)
{
{
elim_int_j = cur_j;
elim_val = fabs(cur_row_val-1.0);
}
......@@ -558,7 +556,7 @@ make_constraints_independent(
// sw.start();
double val = -(*c_it)/elim_val_cur;
add_row_simultaneously( c_it.index(), val, gmm::mat_row(_constraints, i), _constraints, constraints_c);
add_row_simultaneously((int)c_it.index(), val, gmm::mat_row(_constraints, i), _constraints, constraints_c);
// make sure the eliminated entry is 0 on all other rows and not 1e-17
_constraints( c_it.index(), elim_j) = 0;
constraints_c(c_it.index(), elim_j) = 0;
......@@ -586,14 +584,17 @@ make_constraints_independent_reordering(
{
DEB_enter_func;
// setup linear transformation for rhs, start with identity
unsigned int nr = gmm::mat_nrows(_constraints);
gmm::size_type nr = gmm::mat_nrows(_constraints);
gmm::resize(rhs_update_table_.D_, nr, nr);
gmm::clear(rhs_update_table_.D_);
for(unsigned int i=0; i<nr; ++i) rhs_update_table_.D_(i,i) = 1.0;
for(gmm::size_type i=0; i<nr; ++i)
rhs_update_table_.D_(i,i) = 1.0;
// Base::StopWatch sw;
// number of variables
int n_vars = gmm::mat_ncols(_constraints);
// AF: Why was n_vars signed? Can it be zero? Later we subtract 1
const gmm::size_type n_vars = gmm::mat_ncols(_constraints);
// TODO Check: HZ added 14.08.09
_c_elim.clear();
......@@ -601,7 +602,7 @@ make_constraints_independent_reordering(
// build round map
std::vector<bool> roundmap( n_vars, false);
for(unsigned int i=0; i<_idx_to_round.size(); ++i)
for(size_t i=0; i<_idx_to_round.size(); ++i)
roundmap[_idx_to_round[i]] = true;
// copy constraints into column matrix (for faster update via iterators)
......@@ -613,17 +614,19 @@ make_constraints_independent_reordering(
// init priority queue
MutablePriorityQueueT<unsigned int, unsigned int> queue;
queue.clear( nr );
queue.clear( static_cast<int>(nr) );
for(unsigned int i=0; i<nr; ++i)
{
int cur_nnz = gmm::nnz( gmm::mat_row(_constraints,i));
if( _constraints(i,n_vars-1) != 0.0) --cur_nnz;
gmm::size_type cur_nnz = gmm::nnz( gmm::mat_row(_constraints,i));
if( _constraints(i,n_vars-1) != 0.0)
--cur_nnz;
queue.update(i, cur_nnz);
queue.update(i, static_cast<int>(cur_nnz));
}
std::vector<bool> row_visited(nr, false);
std::vector<unsigned int> row_ordering; row_ordering.reserve(nr);
std::vector<gmm::size_type> row_ordering;
row_ordering.reserve(nr);
// for all conditions
......@@ -663,16 +666,16 @@ make_constraints_independent_reordering(
for(; row_it != row_end; ++row_it)
{
int cur_j = row_it.index();
int cur_j = static_cast<int>(row_it.index());
// do not use the constant part
if( cur_j != n_vars - 1 )
if (cur_j != n_vars - 1)
{
// found real valued var? -> finished (UPDATE: no not any more, find biggest real value to avoid x/1e-13)
if( !roundmap[ cur_j ])
if (!roundmap[ cur_j ])
{
if( fabs(*row_it) > max_elim_val)
if (fabs(*row_it) > max_elim_val)
{
elim_j = cur_j;
elim_j = (int)cur_j;
max_elim_val = fabs(*row_it);
}
//break;
......@@ -683,20 +686,19 @@ make_constraints_independent_reordering(
// gcd
// if the coefficient of an integer variable is not an integer, then
// the variable most problably will not be (expect if all coeffs are the same, e.g. 0.5)
if( (double(int(cur_row_val))- cur_row_val) != 0.0)
{
// std::cerr << __FUNCTION__ << " Warning: coefficient of integer variable is NOT integer: "
// << cur_row_val << std::endl;
gcd_update_valid = false;
}
if ((double(int(cur_row_val))- cur_row_val) != 0.0)
{
DEB_warning(2, "coefficient of integer variable is NOT integer : " << cur_row_val);
gcd_update_valid = false;
}
v_gcd[n_ints] = cur_row_val;
v_gcd[n_ints] = static_cast<int>(cur_row_val);
++n_ints;
// store integer closest to 1, must be greater than epsilon_
if( fabs(cur_row_val-1.0) < elim_val && cur_row_val > epsilon_)
{
elim_int_j = cur_j;
elim_int_j = (int)cur_j;
elim_val = fabs(cur_row_val-1.0);
}
}
......@@ -734,7 +736,7 @@ make_constraints_independent_reordering(
if( do_gcd_ && gcd_update_valid)
{
// perform gcd update
bool gcd_ok = update_constraint_gcd( _constraints, i, elim_j, v_gcd, n_ints);
bool gcd_ok = update_constraint_gcd( _constraints, (int)i, elim_j, v_gcd, n_ints);
DEB_warning_if( !gcd_ok && (noisy_ > 0), 1, " GCD update failed! "
<< DEB_os_str(gmm::mat_const_row(_constraints, i)) )
}
......@@ -755,7 +757,7 @@ make_constraints_independent_reordering(
// is this condition dependent?
if( elim_j != -1 )
if (elim_j != -1)
{
// get elim variable value
double elim_val_cur = _constraints(i, elim_j);
......@@ -764,30 +766,34 @@ make_constraints_independent_reordering(
CVector col = constraints_c.col(elim_j);
// iterate over column
typename gmm::linalg_traits<CVector>::const_iterator c_it = gmm::vect_const_begin(col);
typename gmm::linalg_traits<CVector>::const_iterator c_end = gmm::vect_const_end(col);
typename gmm::linalg_traits<CVector>::const_iterator c_it = gmm::vect_const_begin(col);
typename gmm::linalg_traits<CVector>::const_iterator c_end = gmm::vect_const_end(col);
for(; c_it != c_end; ++c_it)
// if( c_it.index() > i)
if( !row_visited[c_it.index()])
for (; c_it != c_end; ++c_it)
{
// if( c_it.index() > i)
if (!row_visited[c_it.index()])
{
// sw.start();
double val = -(*c_it)/elim_val_cur;
add_row_simultaneously( c_it.index(), val, gmm::mat_row(_constraints, i), _constraints, constraints_c);
// sw.start();
double val = -(*c_it) / elim_val_cur;
add_row_simultaneously((int)c_it.index(), val, gmm::mat_row(_constraints, i), _constraints, constraints_c);
// make sure the eliminated entry is 0 on all other rows and not 1e-17
_constraints( c_it.index(), elim_j) = 0;
_constraints(c_it.index(), elim_j) = 0;
constraints_c(c_it.index(), elim_j) = 0;
int cur_idx = c_it.index();
int cur_nnz = gmm::nnz( gmm::mat_row(_constraints,cur_idx));
if( _constraints(cur_idx,n_vars-1) != 0.0) --cur_nnz;
gmm::size_type cur_idx = c_it.index();
gmm::size_type cur_nnz = gmm::nnz( gmm::mat_row(_constraints,cur_idx));
if( _constraints(cur_idx,n_vars-1) != 0.0)
--cur_nnz;
queue.update(cur_idx, cur_nnz);
queue.update(static_cast<int>(cur_idx),
static_cast<int>(cur_nnz));
// update linear transition of rhs
gmm::add(gmm::scaled(gmm::mat_row(rhs_update_table_.D_, i), val),
gmm::mat_row(rhs_update_table_.D_, c_it.index()));
gmm::mat_row(rhs_update_table_.D_, c_it.index()));
}
}
}
}
// // check result
......@@ -857,10 +863,10 @@ update_constraint_gcd( RMatrixT& _constraints,
for( ; row_it!=row_end; ++row_it)
{
int cur_j = row_it.index();
gmm::size_type cur_j = row_it.index();
_constraints(_row_i, cur_j) = (*row_it)/i_gcd;
}
int elim_coeff = abs(_constraints(_row_i, _elim_j));
gmm::size_type elim_coeff = static_cast<gmm::size_type>(std::abs(_constraints(_row_i, _elim_j)));
DEB_error_if( elim_coeff != 1, "elimination coefficient " << elim_coeff
<< " will (most probably) NOT lead to an integer solution!")
return true;
......@@ -1026,7 +1032,7 @@ eliminate_constraints(
//_rhs[con_it.index()] -= cur_rhs * (( *con_it )/cur_val);
// rhs_update_table_.append(con_it.index(), -1.0 * (( *con_it )/cur_val), cur_j);
rhs_update_table_.append(con_it.index(), -1.0 * (( *con_it )/cur_val), cur_j, false);
rhs_update_table_.append(static_cast<int>(con_it.index()), -1.0 * (( *con_it )/cur_val), static_cast<int>(cur_j), false);
//std::cerr << con_it.index() << " += " << -1.0*(( *con_it )/cur_val) << " * " << cur_rhs << " (["<<cur_j<<"] = "<<_rhs[cur_j]<<") " << std::endl;
}
......@@ -1047,7 +1053,7 @@ eliminate_constraints(
_A( col_it.index(), con_it.index() ) -= ( *col_it )*(( *con_it )/cur_val);
//_rhs[col_it.index()] += constraint_rhs*( *col_it )/cur_val;
// rhs_update_table_.append(col_it.index(), constraint_rhs*( *col_it )/cur_val);
rhs_update_table_.append(col_it.index(), -( *col_it )/cur_val, i, true);
rhs_update_table_.append(static_cast<int>(col_it.index()), -( *col_it )/cur_val, i, true);
//std::cerr << col_it.index() << " += " << constraint_rhs*( *col_it )/cur_val << std::endl;
}
......@@ -1108,7 +1114,7 @@ eliminate_constraints(
template<class RowT, class MatrixT>
void
ConstrainedSolver::
add_row( int _row_i,
add_row( gmm::size_type _row_i,
double _coeff,
RowT _row,
MatrixT& _mat )
......@@ -1128,11 +1134,11 @@ add_row( int _row_i,
template<class RowT, class RMatrixT, class CMatrixT>
void
ConstrainedSolver::
add_row_simultaneously( int _row_i,
double _coeff,
RowT _row,
RMatrixT& _rmat,
CMatrixT& _cmat )
add_row_simultaneously( gmm::size_type _row_i,
double _coeff,
RowT _row,
RMatrixT& _rmat,
CMatrixT& _cmat )
{
typedef typename gmm::linalg_traits<RowT>::const_iterator RIter;
RIter r_it = gmm::vect_const_begin(_row);
......@@ -1248,7 +1254,7 @@ restore_eliminated_vars( RMatrixT& _constraints,
_x.back() = 1.0;
// reverse iterate from prelast element
for(int i=_new_idx.size()-2; i>= 0; --i)
for(int i= static_cast<int>(_new_idx.size())-2; i>= 0; --i) // AF: Can this be negative?
{
if( _new_idx[i] != -1)
{
......@@ -1259,7 +1265,7 @@ restore_eliminated_vars( RMatrixT& _constraints,
}
// reverse iterate
for(int i=_c_elim.size()-1; i>=0; --i)
for(int i = static_cast<int>(_c_elim.size())-1; i>=0; --i) // AF: Can this be negative?
{
int cur_var = _c_elim[i];
......
Solver/EigenLDLTSolver.hh
View file @ dea8257e
......@@ -45,9 +45,12 @@
#include <vector>
#include <Base/Code/Quality.hh>
LOW_CODE_QUALITY_SECTION_BEGIN
#include <Eigen/Eigen>
#include <Eigen/Sparse>
#include <Eigen/SparseCholesky>
LOW_CODE_QUALITY_SECTION_END
//== NAMESPACES ===============================================================
......@@ -55,44 +58,46 @@
namespace COMISO {
//== CLASS DEFINITION =========================================================
class COMISODLLEXPORT EigenLDLTSolver
class EigenLDLTSolver
{
public:
// _size is maximal size this instance can handle (smaller problems are possible!!!)
EigenLDLTSolver();
~EigenLDLTSolver();
COMISODLLEXPORT EigenLDLTSolver();
COMISODLLEXPORT ~EigenLDLTSolver();
COMISODLLEXPORT
bool calc_system( const std::vector<int>& _colptr,
const std::vector<int>& _rowind,
const std::vector<double>& _values );
template< class GMM_MatrixT>
bool calc_system_gmm( const GMM_MatrixT& _mat);
template< class Eigen_MatrixT>
bool calc_system_eigen( const Eigen_MatrixT& _mat);
COMISODLLEXPORT
bool update_system( const std::vector<int>& _colptr,
const std::vector<int>& _rowind,
const std::vector<double>& _values );
template< class GMM_MatrixT>
bool update_system_gmm( const GMM_MatrixT& _mat);
template< class Eigen_MatrixT>
bool update_system_eigen( const Eigen_MatrixT& _mat);
COMISODLLEXPORT
bool solve ( double * _x0, double * _b);
COMISODLLEXPORT
bool solve ( std::vector<double>& _x0, std::vector<double>& _b);
COMISODLLEXPORT
bool& show_timings();
COMISODLLEXPORT
int dimension();
private:
......
Solver/Eigen_Tools.hh
View file @ dea8257e
......@@ -33,16 +33,18 @@
//== INCLUDES =================================================================
#include <Base/Code/Quality.hh>
#include <iostream>
#include <vector>
#include <algorithm>
#include <limits>
#include <cmath>
LOW_CODE_QUALITY_SECTION_BEGIN
#include <Eigen/Dense>
#define EIGEN_YES_I_KNOW_SPARSE_MODULE_IS_NOT_STABLE_YET
#include <Eigen/Sparse>
LOW_CODE_QUALITY_SECTION_END
#if COMISO_SUITESPARSE_AVAILABLE
#include <cholmod.h>
......
Solver/Eigen_ToolsT.cc
View file @ dea8257e
......@@ -340,7 +340,7 @@ void cholmod_to_eigen( const cholmod_sparse& _AC, MatrixT& _A)
typedef Eigen::Triplet< Scalar > Triplet;
size_t nzmax( _AC.nzmax);
std::cerr << __FUNCTION__ << " row " << _AC.nrow << " col " << _AC.ncol << " stype " << _AC.stype << std::endl;
_A = MatrixT(_AC.nrow, _AC.ncol);
_A = MatrixT((int)_AC.nrow, (int)_AC.ncol);
std::vector< Triplet > triplets;
triplets.reserve(nzmax);
......@@ -365,7 +365,7 @@ void cholmod_to_eigen( const cholmod_sparse& _AC, MatrixT& _A)
for(SuiteSparse_long i=0; i<(SuiteSparse_long)_AC.ncol; ++i)
for(SuiteSparse_long j= P[i]; j< P[i+1]; ++j)
//_A( I[j], i) += X[j]; // += really needed?
triplets.push_back( Triplet( I[j], i, X[j]));
triplets.push_back(Triplet((int)I[j], (int)i, X[j]));
}
else
{
......@@ -374,7 +374,7 @@ void cholmod_to_eigen( const cholmod_sparse& _AC, MatrixT& _A)
for(int i=0; i<(int)_AC.ncol; ++i)
for(int j= P[i]; j< P[i+1]; ++j)
triplets.push_back( Triplet( I[j], i, X[j]));
triplets.push_back(Triplet((int)I[j], (int)i, X[j]));
//_A( I[j], i) += X[j];
}
......@@ -391,11 +391,11 @@ void cholmod_to_eigen( const cholmod_sparse& _AC, MatrixT& _A)
for(SuiteSparse_long j=P[i]; j<P[i+1]; ++j)
{
//_A(I[j], i) += X[j];
triplets.push_back( Triplet( I[j], i, X[j]));
triplets.push_back(Triplet((int)I[j], (int)i, X[j]));
// add up symmetric part
if( I[j] != i)
triplets.push_back( Triplet( i, I[j], X[j]));
triplets.push_back(Triplet((int)i, (int)I[j], X[j]));
//_A(i,I[j]) += X[j];
}
}
......@@ -731,19 +731,21 @@ void f(const gmm::row_matrix<GMM_VectorT>& _G, EIGEN_MatrixT& _E)
std::vector< Triplet > triplets;
triplets.reserve(gmm::nnz(_G));
for(unsigned int i=0; i<gmm::mat_nrows(_G); ++i)
for(gmm::size_type i=0; i<gmm::mat_nrows(_G); ++i)
{
RowT row = mat_const_row( _G, i );
CIter it = gmm::vect_const_begin( row );
CIter ite = gmm::vect_const_end( row );
for ( ; it!=ite; ++it )
triplets.push_back( Triplet( i, it.index(), *it));
triplets.push_back(
Triplet( static_cast<int>(i), static_cast<int>(it.index()), *it));
}
// generate eigen matrix
_E = EIGEN_MatrixT( gmm::mat_nrows(_G), gmm::mat_ncols(_G));
_E = EIGEN_MatrixT( static_cast<int>(gmm::mat_nrows(_G)),
static_cast<int>(gmm::mat_ncols(_G)));
_E.setFromTriplets( triplets.begin(), triplets.end());
}
......@@ -769,11 +771,12 @@ void f(const gmm::csc_matrix<GMM_RealT,0>& _G, EIGEN_MatrixT& _E)
CIter it = gmm::vect_const_begin( col );
CIter ite = gmm::vect_const_end( col );
for ( ; it!=ite; ++it )
triplets.push_back( Triplet( it.index(), i, *it));
triplets.push_back( Triplet( static_cast<int>(it.index()), i, *it));
}
// generate eigen matrix
_E = EIGEN_MatrixT( gmm::mat_nrows(_G), gmm::mat_ncols(_G));
_E = EIGEN_MatrixT( static_cast<int>(gmm::mat_nrows(_G)),
static_cast<int>( gmm::mat_ncols(_G)));
_E.setFromTriplets( triplets.begin(), triplets.end());
}
......
Solver/GMM_ToolsT.cc
View file @ dea8257e
......@@ -65,13 +65,13 @@ void eliminate_csc_vars(
typedef unsigned int uint;
typedef typename gmm::csc_matrix<RealT> MatrixT;
unsigned int nc = _A.nc;
unsigned int nr = _A.nr;
gmm::size_type nc = _A.nc;
gmm::size_type nr = _A.nr;
unsigned int n_new = nc - _evar.size();
gmm::size_type n_new = nc - _evar.size();
// modify rhs
for ( unsigned int k=0; k<_evar.size(); ++k )
for ( gmm::size_type k=0; k<_evar.size(); ++k )
{
IntegerT i = _evar[k];
......@@ -95,7 +95,7 @@ void eliminate_csc_vars(
// build subindex set and update rhs
std::vector<size_t> si( n_new );
int cur_evar_idx=0;
for ( unsigned int i=0; i<nc; ++i )
for ( gmm::size_type i=0; i<nc; ++i )
{
unsigned int next_i = evar[cur_evar_idx];
if ( i != next_i )
......@@ -119,7 +119,7 @@ void eliminate_csc_vars(