• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // Small bench routine for Eigen available in Eigen
2 // (C) Desire NUENTSA WAKAM, INRIA
3 
4 #include <iostream>
5 #include <fstream>
6 #include <iomanip>
7 #include <Eigen/Jacobi>
8 #include <Eigen/Householder>
9 #include <Eigen/IterativeLinearSolvers>
10 #include <Eigen/LU>
11 #include <unsupported/Eigen/SparseExtra>
12 //#include <Eigen/SparseLU>
13 #include <Eigen/SuperLUSupport>
14 // #include <unsupported/Eigen/src/IterativeSolvers/Scaling.h>
15 #include <bench/BenchTimer.h>
16 #include <unsupported/Eigen/IterativeSolvers>
17 using namespace std;
18 using namespace Eigen;
19 
main(int argc,char ** args)20 int main(int argc, char **args)
21 {
22   SparseMatrix<double, ColMajor> A;
23   typedef SparseMatrix<double, ColMajor>::Index Index;
24   typedef Matrix<double, Dynamic, Dynamic> DenseMatrix;
25   typedef Matrix<double, Dynamic, 1> DenseRhs;
26   VectorXd b, x, tmp;
27   BenchTimer timer,totaltime;
28   //SparseLU<SparseMatrix<double, ColMajor> >   solver;
29 //   SuperLU<SparseMatrix<double, ColMajor> >   solver;
30   ConjugateGradient<SparseMatrix<double, ColMajor>, Lower,IncompleteCholesky<double,Lower> > solver;
31   ifstream matrix_file;
32   string line;
33   int  n;
34   // Set parameters
35 //   solver.iparm(IPARM_THREAD_NBR) = 4;
36   /* Fill the matrix with sparse matrix stored in Matrix-Market coordinate column-oriented format */
37   if (argc < 2) assert(false && "please, give the matrix market file ");
38 
39   timer.start();
40   totaltime.start();
41   loadMarket(A, args[1]);
42   cout << "End charging matrix " << endl;
43   bool iscomplex=false, isvector=false;
44   int sym;
45   getMarketHeader(args[1], sym, iscomplex, isvector);
46   if (iscomplex) { cout<< " Not for complex matrices \n"; return -1; }
47   if (isvector) { cout << "The provided file is not a matrix file\n"; return -1;}
48   if (sym != 0) { // symmetric matrices, only the lower part is stored
49     SparseMatrix<double, ColMajor> temp;
50     temp = A;
51     A = temp.selfadjointView<Lower>();
52   }
53   timer.stop();
54 
55   n = A.cols();
56   // ====== TESTS FOR SPARSE TUTORIAL ======
57 //   cout<< "OuterSize " << A.outerSize() << " inner " << A.innerSize() << endl;
58 //   SparseMatrix<double, RowMajor> mat1(A);
59 //   SparseMatrix<double, RowMajor> mat2;
60 //   cout << " norm of A " << mat1.norm() << endl; ;
61 //   PermutationMatrix<Dynamic, Dynamic, int> perm(n);
62 //   perm.resize(n,1);
63 //   perm.indices().setLinSpaced(n, 0, n-1);
64 //   mat2 = perm * mat1;
65 //   mat.subrows();
66 //   mat2.resize(n,n);
67 //   mat2.reserve(10);
68 //   mat2.setConstant();
69 //   std::cout<< "NORM " << mat1.squaredNorm()<< endl;
70 
71   cout<< "Time to load the matrix " << timer.value() <<endl;
72   /* Fill the right hand side */
73 
74 //   solver.set_restart(374);
75   if (argc > 2)
76     loadMarketVector(b, args[2]);
77   else
78   {
79     b.resize(n);
80     tmp.resize(n);
81 //       tmp.setRandom();
82     for (int i = 0; i < n; i++) tmp(i) = i;
83     b = A * tmp ;
84   }
85 //   Scaling<SparseMatrix<double> > scal;
86 //   scal.computeRef(A);
87 //   b = scal.LeftScaling().cwiseProduct(b);
88 
89   /* Compute the factorization */
90   cout<< "Starting the factorization "<< endl;
91   timer.reset();
92   timer.start();
93   cout<< "Size of Input Matrix "<< b.size()<<"\n\n";
94   cout<< "Rows and columns "<< A.rows() <<" " <<A.cols() <<"\n";
95   solver.compute(A);
96 //   solver.analyzePattern(A);
97 //   solver.factorize(A);
98   if (solver.info() != Success) {
99     std::cout<< "The solver failed \n";
100     return -1;
101   }
102   timer.stop();
103   float time_comp = timer.value();
104   cout <<" Compute Time " << time_comp<< endl;
105 
106   timer.reset();
107   timer.start();
108   x = solver.solve(b);
109 //   x = scal.RightScaling().cwiseProduct(x);
110   timer.stop();
111   float time_solve = timer.value();
112   cout<< " Time to solve " << time_solve << endl;
113 
114   /* Check the accuracy */
115   VectorXd tmp2 = b - A*x;
116   double tempNorm = tmp2.norm()/b.norm();
117   cout << "Relative norm of the computed solution : " << tempNorm <<"\n";
118 //   cout << "Iterations : " << solver.iterations() << "\n";
119 
120   totaltime.stop();
121   cout << "Total time " << totaltime.value() << "\n";
122 //  std::cout<<x.transpose()<<"\n";
123 
124   return 0;
125 }