xref: /aosp_15_r20/external/eigen/test/product_extra.cpp (revision bf2c37156dfe67e5dfebd6d394bad8b2ab5804d4)
1 // This file is part of Eigen, a lightweight C++ template library
2 // for linear algebra.
3 //
4 // Copyright (C) 2006-2008 Benoit Jacob <[email protected]>
5 //
6 // This Source Code Form is subject to the terms of the Mozilla
7 // Public License v. 2.0. If a copy of the MPL was not distributed
8 // with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
9 
10 #include "main.h"
11 
product_extra(const MatrixType & m)12 template<typename MatrixType> void product_extra(const MatrixType& m)
13 {
14   typedef typename MatrixType::Scalar Scalar;
15   typedef Matrix<Scalar, 1, Dynamic> RowVectorType;
16   typedef Matrix<Scalar, Dynamic, 1> ColVectorType;
17   typedef Matrix<Scalar, Dynamic, Dynamic,
18                          MatrixType::Flags&RowMajorBit> OtherMajorMatrixType;
19 
20   Index rows = m.rows();
21   Index cols = m.cols();
22 
23   MatrixType m1 = MatrixType::Random(rows, cols),
24              m2 = MatrixType::Random(rows, cols),
25              m3(rows, cols),
26              mzero = MatrixType::Zero(rows, cols),
27              identity = MatrixType::Identity(rows, rows),
28              square = MatrixType::Random(rows, rows),
29              res = MatrixType::Random(rows, rows),
30              square2 = MatrixType::Random(cols, cols),
31              res2 = MatrixType::Random(cols, cols);
32   RowVectorType v1 = RowVectorType::Random(rows), vrres(rows);
33   ColVectorType vc2 = ColVectorType::Random(cols), vcres(cols);
34   OtherMajorMatrixType tm1 = m1;
35 
36   Scalar s1 = internal::random<Scalar>(),
37          s2 = internal::random<Scalar>(),
38          s3 = internal::random<Scalar>();
39 
40   VERIFY_IS_APPROX(m3.noalias() = m1 * m2.adjoint(),                 m1 * m2.adjoint().eval());
41   VERIFY_IS_APPROX(m3.noalias() = m1.adjoint() * square.adjoint(),   m1.adjoint().eval() * square.adjoint().eval());
42   VERIFY_IS_APPROX(m3.noalias() = m1.adjoint() * m2,                 m1.adjoint().eval() * m2);
43   VERIFY_IS_APPROX(m3.noalias() = (s1 * m1.adjoint()) * m2,          (s1 * m1.adjoint()).eval() * m2);
44   VERIFY_IS_APPROX(m3.noalias() = ((s1 * m1).adjoint()) * m2,        (numext::conj(s1) * m1.adjoint()).eval() * m2);
45   VERIFY_IS_APPROX(m3.noalias() = (- m1.adjoint() * s1) * (s3 * m2), (- m1.adjoint()  * s1).eval() * (s3 * m2).eval());
46   VERIFY_IS_APPROX(m3.noalias() = (s2 * m1.adjoint() * s1) * m2,     (s2 * m1.adjoint()  * s1).eval() * m2);
47   VERIFY_IS_APPROX(m3.noalias() = (-m1*s2) * s1*m2.adjoint(),        (-m1*s2).eval() * (s1*m2.adjoint()).eval());
48 
49   // a very tricky case where a scale factor has to be automatically conjugated:
50   VERIFY_IS_APPROX( m1.adjoint() * (s1*m2).conjugate(), (m1.adjoint()).eval() * ((s1*m2).conjugate()).eval());
51 
52 
53   // test all possible conjugate combinations for the four matrix-vector product cases:
54 
55   VERIFY_IS_APPROX((-m1.conjugate() * s2) * (s1 * vc2),
56                    (-m1.conjugate()*s2).eval() * (s1 * vc2).eval());
57   VERIFY_IS_APPROX((-m1 * s2) * (s1 * vc2.conjugate()),
58                    (-m1*s2).eval() * (s1 * vc2.conjugate()).eval());
59   VERIFY_IS_APPROX((-m1.conjugate() * s2) * (s1 * vc2.conjugate()),
60                    (-m1.conjugate()*s2).eval() * (s1 * vc2.conjugate()).eval());
61 
62   VERIFY_IS_APPROX((s1 * vc2.transpose()) * (-m1.adjoint() * s2),
63                    (s1 * vc2.transpose()).eval() * (-m1.adjoint()*s2).eval());
64   VERIFY_IS_APPROX((s1 * vc2.adjoint()) * (-m1.transpose() * s2),
65                    (s1 * vc2.adjoint()).eval() * (-m1.transpose()*s2).eval());
66   VERIFY_IS_APPROX((s1 * vc2.adjoint()) * (-m1.adjoint() * s2),
67                    (s1 * vc2.adjoint()).eval() * (-m1.adjoint()*s2).eval());
68 
69   VERIFY_IS_APPROX((-m1.adjoint() * s2) * (s1 * v1.transpose()),
70                    (-m1.adjoint()*s2).eval() * (s1 * v1.transpose()).eval());
71   VERIFY_IS_APPROX((-m1.transpose() * s2) * (s1 * v1.adjoint()),
72                    (-m1.transpose()*s2).eval() * (s1 * v1.adjoint()).eval());
73   VERIFY_IS_APPROX((-m1.adjoint() * s2) * (s1 * v1.adjoint()),
74                    (-m1.adjoint()*s2).eval() * (s1 * v1.adjoint()).eval());
75 
76   VERIFY_IS_APPROX((s1 * v1) * (-m1.conjugate() * s2),
77                    (s1 * v1).eval() * (-m1.conjugate()*s2).eval());
78   VERIFY_IS_APPROX((s1 * v1.conjugate()) * (-m1 * s2),
79                    (s1 * v1.conjugate()).eval() * (-m1*s2).eval());
80   VERIFY_IS_APPROX((s1 * v1.conjugate()) * (-m1.conjugate() * s2),
81                    (s1 * v1.conjugate()).eval() * (-m1.conjugate()*s2).eval());
82 
83   VERIFY_IS_APPROX((-m1.adjoint() * s2) * (s1 * v1.adjoint()),
84                    (-m1.adjoint()*s2).eval() * (s1 * v1.adjoint()).eval());
85 
86   // test the vector-matrix product with non aligned starts
87   Index i = internal::random<Index>(0,m1.rows()-2);
88   Index j = internal::random<Index>(0,m1.cols()-2);
89   Index r = internal::random<Index>(1,m1.rows()-i);
90   Index c = internal::random<Index>(1,m1.cols()-j);
91   Index i2 = internal::random<Index>(0,m1.rows()-1);
92   Index j2 = internal::random<Index>(0,m1.cols()-1);
93 
94   VERIFY_IS_APPROX(m1.col(j2).adjoint() * m1.block(0,j,m1.rows(),c), m1.col(j2).adjoint().eval() * m1.block(0,j,m1.rows(),c).eval());
95   VERIFY_IS_APPROX(m1.block(i,0,r,m1.cols()) * m1.row(i2).adjoint(), m1.block(i,0,r,m1.cols()).eval() * m1.row(i2).adjoint().eval());
96 
97   // test negative strides
98   {
99     Map<MatrixType,Unaligned,Stride<Dynamic,Dynamic> > map1(&m1(rows-1,cols-1), rows, cols, Stride<Dynamic,Dynamic>(-m1.outerStride(),-1));
100     Map<MatrixType,Unaligned,Stride<Dynamic,Dynamic> > map2(&m2(rows-1,cols-1), rows, cols, Stride<Dynamic,Dynamic>(-m2.outerStride(),-1));
101     Map<RowVectorType,Unaligned,InnerStride<-1> > mapv1(&v1(v1.size()-1), v1.size(), InnerStride<-1>(-1));
102     Map<ColVectorType,Unaligned,InnerStride<-1> > mapvc2(&vc2(vc2.size()-1), vc2.size(), InnerStride<-1>(-1));
103     VERIFY_IS_APPROX(MatrixType(map1), m1.reverse());
104     VERIFY_IS_APPROX(MatrixType(map2), m2.reverse());
105     VERIFY_IS_APPROX(m3.noalias() = MatrixType(map1) * MatrixType(map2).adjoint(), m1.reverse() * m2.reverse().adjoint());
106     VERIFY_IS_APPROX(m3.noalias() = map1 * map2.adjoint(), m1.reverse() * m2.reverse().adjoint());
107     VERIFY_IS_APPROX(map1 * vc2, m1.reverse() * vc2);
108     VERIFY_IS_APPROX(m1 * mapvc2, m1 * mapvc2);
109     VERIFY_IS_APPROX(map1.adjoint() * v1.transpose(), m1.adjoint().reverse() * v1.transpose());
110     VERIFY_IS_APPROX(m1.adjoint() * mapv1.transpose(), m1.adjoint() * v1.reverse().transpose());
111   }
112 
113   // regression test
114   MatrixType tmp = m1 * m1.adjoint() * s1;
115   VERIFY_IS_APPROX(tmp, m1 * m1.adjoint() * s1);
116 
117   // regression test for bug 1343, assignment to arrays
118   Array<Scalar,Dynamic,1> a1 = m1 * vc2;
119   VERIFY_IS_APPROX(a1.matrix(),m1*vc2);
120   Array<Scalar,Dynamic,1> a2 = s1 * (m1 * vc2);
121   VERIFY_IS_APPROX(a2.matrix(),s1*m1*vc2);
122   Array<Scalar,1,Dynamic> a3 = v1 * m1;
123   VERIFY_IS_APPROX(a3.matrix(),v1*m1);
124   Array<Scalar,Dynamic,Dynamic> a4 = m1 * m2.adjoint();
125   VERIFY_IS_APPROX(a4.matrix(),m1*m2.adjoint());
126 }
127 
128 // Regression test for bug reported at http://forum.kde.org/viewtopic.php?f=74&t=96947
mat_mat_scalar_scalar_product()129 void mat_mat_scalar_scalar_product()
130 {
131   Eigen::Matrix2Xd dNdxy(2, 3);
132   dNdxy << -0.5, 0.5, 0,
133            -0.3, 0, 0.3;
134   double det = 6.0, wt = 0.5;
135   VERIFY_IS_APPROX(dNdxy.transpose()*dNdxy*det*wt, det*wt*dNdxy.transpose()*dNdxy);
136 }
137 
138 template <typename MatrixType>
zero_sized_objects(const MatrixType & m)139 void zero_sized_objects(const MatrixType& m)
140 {
141   typedef typename MatrixType::Scalar Scalar;
142   const int PacketSize  = internal::packet_traits<Scalar>::size;
143   const int PacketSize1 = PacketSize>1 ?  PacketSize-1 : 1;
144   Index rows = m.rows();
145   Index cols = m.cols();
146 
147   {
148     MatrixType res, a(rows,0), b(0,cols);
149     VERIFY_IS_APPROX( (res=a*b), MatrixType::Zero(rows,cols) );
150     VERIFY_IS_APPROX( (res=a*a.transpose()), MatrixType::Zero(rows,rows) );
151     VERIFY_IS_APPROX( (res=b.transpose()*b), MatrixType::Zero(cols,cols) );
152     VERIFY_IS_APPROX( (res=b.transpose()*a.transpose()), MatrixType::Zero(cols,rows) );
153   }
154 
155   {
156     MatrixType res, a(rows,cols), b(cols,0);
157     res = a*b;
158     VERIFY(res.rows()==rows && res.cols()==0);
159     b.resize(0,rows);
160     res = b*a;
161     VERIFY(res.rows()==0 && res.cols()==cols);
162   }
163 
164   {
165     Matrix<Scalar,PacketSize,0> a;
166     Matrix<Scalar,0,1> b;
167     Matrix<Scalar,PacketSize,1> res;
168     VERIFY_IS_APPROX( (res=a*b), MatrixType::Zero(PacketSize,1) );
169     VERIFY_IS_APPROX( (res=a.lazyProduct(b)), MatrixType::Zero(PacketSize,1) );
170   }
171 
172   {
173     Matrix<Scalar,PacketSize1,0> a;
174     Matrix<Scalar,0,1> b;
175     Matrix<Scalar,PacketSize1,1> res;
176     VERIFY_IS_APPROX( (res=a*b), MatrixType::Zero(PacketSize1,1) );
177     VERIFY_IS_APPROX( (res=a.lazyProduct(b)), MatrixType::Zero(PacketSize1,1) );
178   }
179 
180   {
181     Matrix<Scalar,PacketSize,Dynamic> a(PacketSize,0);
182     Matrix<Scalar,Dynamic,1> b(0,1);
183     Matrix<Scalar,PacketSize,1> res;
184     VERIFY_IS_APPROX( (res=a*b), MatrixType::Zero(PacketSize,1) );
185     VERIFY_IS_APPROX( (res=a.lazyProduct(b)), MatrixType::Zero(PacketSize,1) );
186   }
187 
188   {
189     Matrix<Scalar,PacketSize1,Dynamic> a(PacketSize1,0);
190     Matrix<Scalar,Dynamic,1> b(0,1);
191     Matrix<Scalar,PacketSize1,1> res;
192     VERIFY_IS_APPROX( (res=a*b), MatrixType::Zero(PacketSize1,1) );
193     VERIFY_IS_APPROX( (res=a.lazyProduct(b)), MatrixType::Zero(PacketSize1,1) );
194   }
195 }
196 
197 template<int>
bug_127()198 void bug_127()
199 {
200   // Bug 127
201   //
202   // a product of the form lhs*rhs with
203   //
204   // lhs:
205   // rows = 1, cols = 4
206   // RowsAtCompileTime = 1, ColsAtCompileTime = -1
207   // MaxRowsAtCompileTime = 1, MaxColsAtCompileTime = 5
208   //
209   // rhs:
210   // rows = 4, cols = 0
211   // RowsAtCompileTime = -1, ColsAtCompileTime = -1
212   // MaxRowsAtCompileTime = 5, MaxColsAtCompileTime = 1
213   //
214   // was failing on a runtime assertion, because it had been mis-compiled as a dot product because Product.h was using the
215   // max-sizes to detect size 1 indicating vectors, and that didn't account for 0-sized object with max-size 1.
216 
217   Matrix<float,1,Dynamic,RowMajor,1,5> a(1,4);
218   Matrix<float,Dynamic,Dynamic,ColMajor,5,1> b(4,0);
219   a*b;
220 }
221 
bug_817()222 template<int> void bug_817()
223 {
224   ArrayXXf B = ArrayXXf::Random(10,10), C;
225   VectorXf x = VectorXf::Random(10);
226   C = (x.transpose()*B.matrix());
227   B = (x.transpose()*B.matrix());
228   VERIFY_IS_APPROX(B,C);
229 }
230 
231 template<int>
unaligned_objects()232 void unaligned_objects()
233 {
234   // Regression test for the bug reported here:
235   // http://forum.kde.org/viewtopic.php?f=74&t=107541
236   // Recall the matrix*vector kernel avoid unaligned loads by loading two packets and then reassemble then.
237   // There was a mistake in the computation of the valid range for fully unaligned objects: in some rare cases,
238   // memory was read outside the allocated matrix memory. Though the values were not used, this might raise segfault.
239   for(int m=450;m<460;++m)
240   {
241     for(int n=8;n<12;++n)
242     {
243       MatrixXf M(m, n);
244       VectorXf v1(n), r1(500);
245       RowVectorXf v2(m), r2(16);
246 
247       M.setRandom();
248       v1.setRandom();
249       v2.setRandom();
250       for(int o=0; o<4; ++o)
251       {
252         r1.segment(o,m).noalias() = M * v1;
253         VERIFY_IS_APPROX(r1.segment(o,m), M * MatrixXf(v1));
254         r2.segment(o,n).noalias() = v2 * M;
255         VERIFY_IS_APPROX(r2.segment(o,n), MatrixXf(v2) * M);
256       }
257     }
258   }
259 }
260 
261 template<typename T>
262 EIGEN_DONT_INLINE
test_compute_block_size(Index m,Index n,Index k)263 Index test_compute_block_size(Index m, Index n, Index k)
264 {
265   Index mc(m), nc(n), kc(k);
266   internal::computeProductBlockingSizes<T,T>(kc, mc, nc);
267   return kc+mc+nc;
268 }
269 
270 template<typename T>
compute_block_size()271 Index compute_block_size()
272 {
273   Index ret = 0;
274   ret += test_compute_block_size<T>(0,1,1);
275   ret += test_compute_block_size<T>(1,0,1);
276   ret += test_compute_block_size<T>(1,1,0);
277   ret += test_compute_block_size<T>(0,0,1);
278   ret += test_compute_block_size<T>(0,1,0);
279   ret += test_compute_block_size<T>(1,0,0);
280   ret += test_compute_block_size<T>(0,0,0);
281   return ret;
282 }
283 
284 template<typename>
aliasing_with_resize()285 void aliasing_with_resize()
286 {
287   Index m = internal::random<Index>(10,50);
288   Index n = internal::random<Index>(10,50);
289   MatrixXd A, B, C(m,n), D(m,m);
290   VectorXd a, b, c(n);
291   C.setRandom();
292   D.setRandom();
293   c.setRandom();
294   double s = internal::random<double>(1,10);
295 
296   A = C;
297   B = A * A.transpose();
298   A = A * A.transpose();
299   VERIFY_IS_APPROX(A,B);
300 
301   A = C;
302   B = (A * A.transpose())/s;
303   A = (A * A.transpose())/s;
304   VERIFY_IS_APPROX(A,B);
305 
306   A = C;
307   B = (A * A.transpose()) + D;
308   A = (A * A.transpose()) + D;
309   VERIFY_IS_APPROX(A,B);
310 
311   A = C;
312   B = D + (A * A.transpose());
313   A = D + (A * A.transpose());
314   VERIFY_IS_APPROX(A,B);
315 
316   A = C;
317   B = s * (A * A.transpose());
318   A = s * (A * A.transpose());
319   VERIFY_IS_APPROX(A,B);
320 
321   A = C;
322   a = c;
323   b = (A * a)/s;
324   a = (A * a)/s;
325   VERIFY_IS_APPROX(a,b);
326 }
327 
328 template<int>
bug_1308()329 void bug_1308()
330 {
331   int n = 10;
332   MatrixXd r(n,n);
333   VectorXd v = VectorXd::Random(n);
334   r = v * RowVectorXd::Ones(n);
335   VERIFY_IS_APPROX(r, v.rowwise().replicate(n));
336   r = VectorXd::Ones(n) * v.transpose();
337   VERIFY_IS_APPROX(r, v.rowwise().replicate(n).transpose());
338 
339   Matrix4d ones44 = Matrix4d::Ones();
340   Matrix4d m44 = Matrix4d::Ones() * Matrix4d::Ones();
341   VERIFY_IS_APPROX(m44,Matrix4d::Constant(4));
342   VERIFY_IS_APPROX(m44.noalias()=ones44*Matrix4d::Ones(), Matrix4d::Constant(4));
343   VERIFY_IS_APPROX(m44.noalias()=ones44.transpose()*Matrix4d::Ones(), Matrix4d::Constant(4));
344   VERIFY_IS_APPROX(m44.noalias()=Matrix4d::Ones()*ones44, Matrix4d::Constant(4));
345   VERIFY_IS_APPROX(m44.noalias()=Matrix4d::Ones()*ones44.transpose(), Matrix4d::Constant(4));
346 
347   typedef Matrix<double,4,4,RowMajor> RMatrix4d;
348   RMatrix4d r44 = Matrix4d::Ones() * Matrix4d::Ones();
349   VERIFY_IS_APPROX(r44,Matrix4d::Constant(4));
350   VERIFY_IS_APPROX(r44.noalias()=ones44*Matrix4d::Ones(), Matrix4d::Constant(4));
351   VERIFY_IS_APPROX(r44.noalias()=ones44.transpose()*Matrix4d::Ones(), Matrix4d::Constant(4));
352   VERIFY_IS_APPROX(r44.noalias()=Matrix4d::Ones()*ones44, Matrix4d::Constant(4));
353   VERIFY_IS_APPROX(r44.noalias()=Matrix4d::Ones()*ones44.transpose(), Matrix4d::Constant(4));
354   VERIFY_IS_APPROX(r44.noalias()=ones44*RMatrix4d::Ones(), Matrix4d::Constant(4));
355   VERIFY_IS_APPROX(r44.noalias()=ones44.transpose()*RMatrix4d::Ones(), Matrix4d::Constant(4));
356   VERIFY_IS_APPROX(r44.noalias()=RMatrix4d::Ones()*ones44, Matrix4d::Constant(4));
357   VERIFY_IS_APPROX(r44.noalias()=RMatrix4d::Ones()*ones44.transpose(), Matrix4d::Constant(4));
358 
359 //   RowVector4d r4;
360   m44.setOnes();
361   r44.setZero();
362   VERIFY_IS_APPROX(r44.noalias() += m44.row(0).transpose() * RowVector4d::Ones(), ones44);
363   r44.setZero();
364   VERIFY_IS_APPROX(r44.noalias() += m44.col(0) * RowVector4d::Ones(), ones44);
365   r44.setZero();
366   VERIFY_IS_APPROX(r44.noalias() += Vector4d::Ones() * m44.row(0), ones44);
367   r44.setZero();
368   VERIFY_IS_APPROX(r44.noalias() += Vector4d::Ones() * m44.col(0).transpose(), ones44);
369 }
370 
EIGEN_DECLARE_TEST(product_extra)371 EIGEN_DECLARE_TEST(product_extra)
372 {
373   for(int i = 0; i < g_repeat; i++) {
374     CALL_SUBTEST_1( product_extra(MatrixXf(internal::random<int>(1,EIGEN_TEST_MAX_SIZE), internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) );
375     CALL_SUBTEST_2( product_extra(MatrixXd(internal::random<int>(1,EIGEN_TEST_MAX_SIZE), internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) );
376     CALL_SUBTEST_2( mat_mat_scalar_scalar_product() );
377     CALL_SUBTEST_3( product_extra(MatrixXcf(internal::random<int>(1,EIGEN_TEST_MAX_SIZE/2), internal::random<int>(1,EIGEN_TEST_MAX_SIZE/2))) );
378     CALL_SUBTEST_4( product_extra(MatrixXcd(internal::random<int>(1,EIGEN_TEST_MAX_SIZE/2), internal::random<int>(1,EIGEN_TEST_MAX_SIZE/2))) );
379     CALL_SUBTEST_1( zero_sized_objects(MatrixXf(internal::random<int>(1,EIGEN_TEST_MAX_SIZE), internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) );
380   }
381   CALL_SUBTEST_5( bug_127<0>() );
382   CALL_SUBTEST_5( bug_817<0>() );
383   CALL_SUBTEST_5( bug_1308<0>() );
384   CALL_SUBTEST_6( unaligned_objects<0>() );
385   CALL_SUBTEST_7( compute_block_size<float>() );
386   CALL_SUBTEST_7( compute_block_size<double>() );
387   CALL_SUBTEST_7( compute_block_size<std::complex<double> >() );
388   CALL_SUBTEST_8( aliasing_with_resize<void>() );
389 
390 }
391