sparse_trisolver.cpp
Go to the documentation of this file.
1 
2 //g++ -O3 -g0 -DNDEBUG sparse_product.cpp -I.. -I/home/gael/Coding/LinearAlgebra/mtl4/ -DDENSITY=0.005 -DSIZE=10000 && ./a.out
3 //g++ -O3 -g0 -DNDEBUG sparse_product.cpp -I.. -I/home/gael/Coding/LinearAlgebra/mtl4/ -DDENSITY=0.05 -DSIZE=2000 && ./a.out
4 // -DNOGMM -DNOMTL
5 // -I /home/gael/Coding/LinearAlgebra/CSparse/Include/ /home/gael/Coding/LinearAlgebra/CSparse/Lib/libcsparse.a
6 
7 #ifndef SIZE
8 #define SIZE 10000
9 #endif
10 
11 #ifndef DENSITY
12 #define DENSITY 0.01
13 #endif
14 
15 #ifndef REPEAT
16 #define REPEAT 1
17 #endif
18 
19 #include "BenchSparseUtil.h"
20 
21 #ifndef MINDENSITY
22 #define MINDENSITY 0.0004
23 #endif
24 
25 #ifndef NBTRIES
26 #define NBTRIES 10
27 #endif
28 
29 #define BENCH(X) \
30  timer.reset(); \
31  for (int _j=0; _j<NBTRIES; ++_j) { \
32  timer.start(); \
33  for (int _k=0; _k<REPEAT; ++_k) { \
34  X \
35  } timer.stop(); }
36 
39 
40 void fillMatrix(float density, int rows, int cols, EigenSparseTriMatrix& dst)
41 {
42  dst.startFill(rows*cols*density);
43  for(int j = 0; j < cols; j++)
44  {
45  for(int i = 0; i < j; i++)
46  {
47  Scalar v = (internal::random<float>(0,1) < density) ? internal::random<Scalar>() : 0;
48  if (v!=0)
49  dst.fill(i,j) = v;
50  }
51  dst.fill(j,j) = internal::random<Scalar>();
52  }
53  dst.endFill();
54 }
55 
56 int main(int argc, char *argv[])
57 {
58  int rows = SIZE;
59  int cols = SIZE;
60  float density = DENSITY;
62  #if 1
63  EigenSparseTriMatrix sm1(rows,cols);
65  DenseVector b = DenseVector::Random(cols);
66  DenseVector x = DenseVector::Random(cols);
67 
68  bool densedone = false;
69 
70  for (float density = DENSITY; density>=MINDENSITY; density*=0.5)
71  {
72  EigenSparseTriMatrix sm1(rows, cols);
73  fillMatrix(density, rows, cols, sm1);
74 
75  // dense matrices
76  #ifdef DENSEMATRIX
77  if (!densedone)
78  {
79  densedone = true;
80  std::cout << "Eigen Dense\t" << density*100 << "%\n";
81  DenseMatrix m1(rows,cols);
83  eiToDense(sm1, m1);
84  m2 = m1;
85 
86  BENCH(x = m1.marked<UpperTriangular>().solveTriangular(b);)
87  std::cout << " colmajor^-1 * b:\t" << timer.value() << endl;
88 // std::cerr << x.transpose() << "\n";
89 
90  BENCH(x = m2.marked<UpperTriangular>().solveTriangular(b);)
91  std::cout << " rowmajor^-1 * b:\t" << timer.value() << endl;
92 // std::cerr << x.transpose() << "\n";
93  }
94  #endif
95 
96  // eigen sparse matrices
97  {
98  std::cout << "Eigen sparse\t" << density*100 << "%\n";
99  EigenSparseTriMatrixRow sm2 = sm1;
100 
101  BENCH(x = sm1.solveTriangular(b);)
102  std::cout << " colmajor^-1 * b:\t" << timer.value() << endl;
103 // std::cerr << x.transpose() << "\n";
104 
105  BENCH(x = sm2.solveTriangular(b);)
106  std::cout << " rowmajor^-1 * b:\t" << timer.value() << endl;
107 // std::cerr << x.transpose() << "\n";
108 
109 // x = b;
110 // BENCH(sm1.inverseProductInPlace(x);)
111 // std::cout << " colmajor^-1 * b:\t" << timer.value() << " (inplace)" << endl;
112 // std::cerr << x.transpose() << "\n";
113 //
114 // x = b;
115 // BENCH(sm2.inverseProductInPlace(x);)
116 // std::cout << " rowmajor^-1 * b:\t" << timer.value() << " (inplace)" << endl;
117 // std::cerr << x.transpose() << "\n";
118  }
119 
120 
121 
122  // CSparse
123  #ifdef CSPARSE
124  {
125  std::cout << "CSparse \t" << density*100 << "%\n";
126  cs *m1;
127  eiToCSparse(sm1, m1);
128 
129  BENCH(x = b; if (!cs_lsolve (m1, x.data())){std::cerr << "cs_lsolve failed\n"; break;}; )
130  std::cout << " colmajor^-1 * b:\t" << timer.value() << endl;
131  }
132  #endif
133 
134  // GMM++
135  #ifndef NOGMM
136  {
137  std::cout << "GMM++ sparse\t" << density*100 << "%\n";
138  GmmSparse m1(rows,cols);
139  gmm::csr_matrix<Scalar> m2;
140  eiToGmm(sm1, m1);
141  gmm::copy(m1,m2);
142  std::vector<Scalar> gmmX(cols), gmmB(cols);
143  Map<Matrix<Scalar,Dynamic,1> >(&gmmX[0], cols) = x;
144  Map<Matrix<Scalar,Dynamic,1> >(&gmmB[0], cols) = b;
145 
146  gmmX = gmmB;
147  BENCH(gmm::upper_tri_solve(m1, gmmX, false);)
148  std::cout << " colmajor^-1 * b:\t" << timer.value() << endl;
149 // std::cerr << Map<Matrix<Scalar,Dynamic,1> >(&gmmX[0], cols).transpose() << "\n";
150 
151  gmmX = gmmB;
152  BENCH(gmm::upper_tri_solve(m2, gmmX, false);)
153  timer.stop();
154  std::cout << " rowmajor^-1 * b:\t" << timer.value() << endl;
155 // std::cerr << Map<Matrix<Scalar,Dynamic,1> >(&gmmX[0], cols).transpose() << "\n";
156  }
157  #endif
158 
159  // MTL4
160  #ifndef NOMTL
161  {
162  std::cout << "MTL4\t" << density*100 << "%\n";
163  MtlSparse m1(rows,cols);
164  MtlSparseRowMajor m2(rows,cols);
165  eiToMtl(sm1, m1);
166  m2 = m1;
167  mtl::dense_vector<Scalar> x(rows, 1.0);
168  mtl::dense_vector<Scalar> b(rows, 1.0);
169 
170  BENCH(x = mtl::upper_trisolve(m1,b);)
171  std::cout << " colmajor^-1 * b:\t" << timer.value() << endl;
172 // std::cerr << x << "\n";
173 
174  BENCH(x = mtl::upper_trisolve(m2,b);)
175  std::cout << " rowmajor^-1 * b:\t" << timer.value() << endl;
176 // std::cerr << x << "\n";
177  }
178  #endif
179 
180 
181  std::cout << "\n\n";
182  }
183  #endif
184 
185  #if 0
186  // bench small matrices (in-place versus return bye value)
187  {
188  timer.reset();
189  for (int _j=0; _j<10; ++_j) {
190  Matrix4f m = Matrix4f::Random();
191  Vector4f b = Vector4f::Random();
192  Vector4f x = Vector4f::Random();
193  timer.start();
194  for (int _k=0; _k<1000000; ++_k) {
195  b = m.inverseProduct(b);
196  }
197  timer.stop();
198  }
199  std::cout << "4x4 :\t" << timer.value() << endl;
200  }
201 
202  {
203  timer.reset();
204  for (int _j=0; _j<10; ++_j) {
205  Matrix4f m = Matrix4f::Random();
206  Vector4f b = Vector4f::Random();
207  Vector4f x = Vector4f::Random();
208  timer.start();
209  for (int _k=0; _k<1000000; ++_k) {
210  m.inverseProductInPlace(x);
211  }
212  timer.stop();
213  }
214  std::cout << "4x4 IP :\t" << timer.value() << endl;
215  }
216  #endif
217 
218  return 0;
219 }
220 
Matrix3f m
Matrix< Scalar, Dynamic, Dynamic > DenseMatrix
SCALAR Scalar
Definition: bench_gemm.cpp:33
void eiToMtl(const EigenSparseMatrix &src, MtlSparse &dst)
SparseMatrix< Scalar, UpperTriangular > EigenSparseTriMatrix
Scalar * b
Definition: benchVecAdd.cpp:17
A versatible sparse matrix representation.
Definition: SparseMatrix.h:96
A matrix or vector expression mapping an existing array of data.
Definition: Map.h:94
MatrixType m2(n_dims)
ArrayXcf v
Definition: Cwise_arg.cpp:1
Matrix< Scalar, Dynamic, 1 > DenseVector
void eiToDense(const EigenSparseMatrix &src, DenseMatrix &dst)
Matrix3d m1
Definition: IOFormat.cpp:2
mtl::compressed2D< Scalar, mtl::matrix::parameters< mtl::tag::col_major > > MtlSparse
gmm::csc_matrix< Scalar > GmmSparse
double value(int TIMER=CPU_TIMER) const
Definition: BenchTimer.h:100
#define BENCH(X)
int main(int argc, char *argv[])
int EIGEN_BLAS_FUNC() copy(int *n, RealScalar *px, int *incx, RealScalar *py, int *incy)
Definition: level1_impl.h:29
#define MINDENSITY
#define DENSITY
set noclip points set clip one set noclip two set bar set border lt lw set xdata set ydata set zdata set x2data set y2data set boxwidth set dummy x
void eiToGmm(const EigenSparseMatrix &src, GmmSparse &dst)
void fillMatrix(float density, int rows, int cols, EigenSparseTriMatrix &dst)
static BenchTimer timer
std::ptrdiff_t j
mtl::compressed2D< Scalar, mtl::matrix::parameters< mtl::tag::row_major > > MtlSparseRowMajor
#define SIZE
SparseMatrix< Scalar, RowMajorBit|UpperTriangular > EigenSparseTriMatrixRow


gtsam
Author(s):
autogenerated on Sat May 8 2021 02:44:18