Austin Schuh | 189376f | 2018-12-20 22:11:15 +1100 | [diff] [blame^] | 1 | // This file is part of Eigen, a lightweight C++ template library |
| 2 | // for linear algebra. |
| 3 | // |
| 4 | // Copyright (C) 2013 Gauthier Brun <brun.gauthier@gmail.com> |
| 5 | // Copyright (C) 2013 Nicolas Carre <nicolas.carre@ensimag.fr> |
| 6 | // Copyright (C) 2013 Jean Ceccato <jean.ceccato@ensimag.fr> |
| 7 | // Copyright (C) 2013 Pierre Zoppitelli <pierre.zoppitelli@ensimag.fr> |
| 8 | // |
| 9 | // This Source Code Form is subject to the terms of the Mozilla |
| 10 | // Public License v. 2.0. If a copy of the MPL was not distributed |
| 11 | // with this file, You can obtain one at http://mozilla.org/MPL/2.0/ |
| 12 | |
| 13 | // discard stack allocation as that too bypasses malloc |
| 14 | #define EIGEN_STACK_ALLOCATION_LIMIT 0 |
| 15 | #define EIGEN_RUNTIME_NO_MALLOC |
| 16 | |
| 17 | #include "main.h" |
| 18 | #include <Eigen/SVD> |
| 19 | #include <iostream> |
| 20 | #include <Eigen/LU> |
| 21 | |
| 22 | |
| 23 | #define SVD_DEFAULT(M) BDCSVD<M> |
| 24 | #define SVD_FOR_MIN_NORM(M) BDCSVD<M> |
| 25 | #include "svd_common.h" |
| 26 | |
| 27 | // Check all variants of JacobiSVD |
| 28 | template<typename MatrixType> |
| 29 | void bdcsvd(const MatrixType& a = MatrixType(), bool pickrandom = true) |
| 30 | { |
| 31 | MatrixType m = a; |
| 32 | if(pickrandom) |
| 33 | svd_fill_random(m); |
| 34 | |
| 35 | CALL_SUBTEST(( svd_test_all_computation_options<BDCSVD<MatrixType> >(m, false) )); |
| 36 | } |
| 37 | |
| 38 | template<typename MatrixType> |
| 39 | void bdcsvd_method() |
| 40 | { |
| 41 | enum { Size = MatrixType::RowsAtCompileTime }; |
| 42 | typedef typename MatrixType::RealScalar RealScalar; |
| 43 | typedef Matrix<RealScalar, Size, 1> RealVecType; |
| 44 | MatrixType m = MatrixType::Identity(); |
| 45 | VERIFY_IS_APPROX(m.bdcSvd().singularValues(), RealVecType::Ones()); |
| 46 | VERIFY_RAISES_ASSERT(m.bdcSvd().matrixU()); |
| 47 | VERIFY_RAISES_ASSERT(m.bdcSvd().matrixV()); |
| 48 | VERIFY_IS_APPROX(m.bdcSvd(ComputeFullU|ComputeFullV).solve(m), m); |
| 49 | } |
| 50 | |
| 51 | // compare the Singular values returned with Jacobi and Bdc |
| 52 | template<typename MatrixType> |
| 53 | void compare_bdc_jacobi(const MatrixType& a = MatrixType(), unsigned int computationOptions = 0) |
| 54 | { |
| 55 | MatrixType m = MatrixType::Random(a.rows(), a.cols()); |
| 56 | BDCSVD<MatrixType> bdc_svd(m); |
| 57 | JacobiSVD<MatrixType> jacobi_svd(m); |
| 58 | VERIFY_IS_APPROX(bdc_svd.singularValues(), jacobi_svd.singularValues()); |
| 59 | if(computationOptions & ComputeFullU) VERIFY_IS_APPROX(bdc_svd.matrixU(), jacobi_svd.matrixU()); |
| 60 | if(computationOptions & ComputeThinU) VERIFY_IS_APPROX(bdc_svd.matrixU(), jacobi_svd.matrixU()); |
| 61 | if(computationOptions & ComputeFullV) VERIFY_IS_APPROX(bdc_svd.matrixV(), jacobi_svd.matrixV()); |
| 62 | if(computationOptions & ComputeThinV) VERIFY_IS_APPROX(bdc_svd.matrixV(), jacobi_svd.matrixV()); |
| 63 | } |
| 64 | |
| 65 | void test_bdcsvd() |
| 66 | { |
| 67 | CALL_SUBTEST_3(( svd_verify_assert<BDCSVD<Matrix3f> >(Matrix3f()) )); |
| 68 | CALL_SUBTEST_4(( svd_verify_assert<BDCSVD<Matrix4d> >(Matrix4d()) )); |
| 69 | CALL_SUBTEST_7(( svd_verify_assert<BDCSVD<MatrixXf> >(MatrixXf(10,12)) )); |
| 70 | CALL_SUBTEST_8(( svd_verify_assert<BDCSVD<MatrixXcd> >(MatrixXcd(7,5)) )); |
| 71 | |
| 72 | CALL_SUBTEST_101(( svd_all_trivial_2x2(bdcsvd<Matrix2cd>) )); |
| 73 | CALL_SUBTEST_102(( svd_all_trivial_2x2(bdcsvd<Matrix2d>) )); |
| 74 | |
| 75 | for(int i = 0; i < g_repeat; i++) { |
| 76 | CALL_SUBTEST_3(( bdcsvd<Matrix3f>() )); |
| 77 | CALL_SUBTEST_4(( bdcsvd<Matrix4d>() )); |
| 78 | CALL_SUBTEST_5(( bdcsvd<Matrix<float,3,5> >() )); |
| 79 | |
| 80 | int r = internal::random<int>(1, EIGEN_TEST_MAX_SIZE/2), |
| 81 | c = internal::random<int>(1, EIGEN_TEST_MAX_SIZE/2); |
| 82 | |
| 83 | TEST_SET_BUT_UNUSED_VARIABLE(r) |
| 84 | TEST_SET_BUT_UNUSED_VARIABLE(c) |
| 85 | |
| 86 | CALL_SUBTEST_6(( bdcsvd(Matrix<double,Dynamic,2>(r,2)) )); |
| 87 | CALL_SUBTEST_7(( bdcsvd(MatrixXf(r,c)) )); |
| 88 | CALL_SUBTEST_7(( compare_bdc_jacobi(MatrixXf(r,c)) )); |
| 89 | CALL_SUBTEST_10(( bdcsvd(MatrixXd(r,c)) )); |
| 90 | CALL_SUBTEST_10(( compare_bdc_jacobi(MatrixXd(r,c)) )); |
| 91 | CALL_SUBTEST_8(( bdcsvd(MatrixXcd(r,c)) )); |
| 92 | CALL_SUBTEST_8(( compare_bdc_jacobi(MatrixXcd(r,c)) )); |
| 93 | |
| 94 | // Test on inf/nan matrix |
| 95 | CALL_SUBTEST_7( (svd_inf_nan<BDCSVD<MatrixXf>, MatrixXf>()) ); |
| 96 | CALL_SUBTEST_10( (svd_inf_nan<BDCSVD<MatrixXd>, MatrixXd>()) ); |
| 97 | } |
| 98 | |
| 99 | // test matrixbase method |
| 100 | CALL_SUBTEST_1(( bdcsvd_method<Matrix2cd>() )); |
| 101 | CALL_SUBTEST_3(( bdcsvd_method<Matrix3f>() )); |
| 102 | |
| 103 | // Test problem size constructors |
| 104 | CALL_SUBTEST_7( BDCSVD<MatrixXf>(10,10) ); |
| 105 | |
| 106 | // Check that preallocation avoids subsequent mallocs |
| 107 | // Disbaled because not supported by BDCSVD |
| 108 | // CALL_SUBTEST_9( svd_preallocate<void>() ); |
| 109 | |
| 110 | CALL_SUBTEST_2( svd_underoverflow<void>() ); |
| 111 | } |
| 112 | |