1 /*
2 * Copyright (c) 2011, Michael Lehn
3 *
4 * All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 *
10 * 1) Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2) Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in
14 * the documentation and/or other materials provided with the
15 * distribution.
16 * 3) Neither the name of the FLENS development group nor the names of
17 * its contributors may be used to endorse or promote products derived
18 * from this software without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
23 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
24 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
25 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
26 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
27 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
28 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
29 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
30 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
31 */
32
33 /* Based on
34 *
35 SUBROUTINE DLAUU2( UPLO, N, A, LDA, INFO )
36 *
37 * -- LAPACK auxiliary routine (version 3.3.1) --
38 * -- LAPACK is a software package provided by Univ. of Tennessee, --
39 * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
40 * -- April 2011
41 */
42
43 #ifndef FLENS_LAPACK_GESV_LAUU2_TCC
44 #define FLENS_LAPACK_GESV_LAUU2_TCC 1
45
46 #include <algorithm>
47 #include <flens/blas/blas.h>
48 #include <flens/lapack/lapack.h>
49
50 #include <flens/lapack/interface/include/f77lapack.h>
51
52 namespace flens { namespace lapack {
53
54 //== generic lapack implementation =============================================
55
56 template <typename MA>
57 void
58 lauu2_generic(TrMatrix<MA> &A)
59 {
60 typedef typename TrMatrix<MA>::ElementType ElementType;
61 typedef typename TrMatrix<MA>::IndexType IndexType;
62
63 const ElementType One(1);
64 const IndexType n = A.dim();
65 const Underscore<IndexType> _;
66 //
67 // Quick return if possible
68 //
69 if (n==0) {
70 return;
71 }
72
73 if (A.upLo()==Upper) {
74 //
75 // Compute the product U * U**T.
76 //
77 for (IndexType i=1; i<=n; ++i) {
78 const ElementType a22 = A(i,i);
79 if (i<n) {
80 A(i,i) = A(i,_(i,n)) * A(i,_(i,n));
81
82 const auto range1 = _(1,i-1);
83 const auto range2 = i;
84 const auto range3 = _(i+1,n);
85
86 auto a12 = A(range1,range2);
87 const auto A13 = A(range1,range3);
88 const auto a23 = A(range2,range3);
89 blas::mv(NoTrans, One, A13, a23, a22, a12);
90 } else {
91 A(_,n) *= a22;
92 }
93 }
94 } else {
95 //
96 // Compute the product L**T * L.
97 //
98 for (IndexType i=1; i<=n; ++i) {
99 const ElementType a22 = A(i,i);
100 if (i<n) {
101 A(i,i) = A(_(i,n),i) * A(_(i,n),i);
102
103 const auto range1 = _(1,i-1);
104 const auto range2 = i;
105 const auto range3 = _(i+1,n);
106
107 auto a21 = A(range2,range1);
108 const auto A31 = A(range3,range1);
109 const auto a32 = A(range3,range2);
110
111 blas::mv(Trans, One, A31, a32, a22, a21);
112 } else {
113 A(n,_) *= a22;
114 }
115 }
116 }
117 }
118
119 //== interface for native lapack ===============================================
120
121 #ifdef CHECK_CXXLAPACK
122
123 template <typename MA>
124 void
125 lauu2_native(TrMatrix<MA> &A)
126 {
127 typedef typename TrMatrix<MA>::ElementType ElementType;
128
129 const char UPLO(A.upLo());
130 const INTEGER N = A.dim();
131 const INTEGER LDA = A.leadingDimension();
132 INTEGER INFO;
133
134 if (IsSame<ElementType,double>::value) {
135 LAPACK_DECL(dlauu2)(&UPLO,
136 &N,
137 A.data(),
138 &LDA,
139 &INFO);
140 } else {
141 ASSERT(0);
142 }
143 ASSERT(INFO==0);
144 }
145
146 #endif // CHECK_CXXLAPACK
147
148 //== public interface ==========================================================
149
150 template <typename MA>
151 void
152 lauu2(TrMatrix<MA> &A)
153 {
154 typedef typename TrMatrix<MA>::IndexType IndexType;
155
156 //
157 // Test the input parameters
158 //
159 ASSERT(A.firstRow()==1);
160 ASSERT(A.firstCol()==1);
161
162 # ifdef CHECK_CXXLAPACK
163 //
164 // Make copies of output arguments
165 //
166 typename TrMatrix<MA>::NoView A_org = A;
167 # endif
168
169 //
170 // Call implementation
171 //
172 lauu2_generic(A);
173
174 # ifdef CHECK_CXXLAPACK
175 //
176 // Make copies of generic results
177 //
178 typename TrMatrix<MA>::NoView A_generic = A;
179 //
180 // Restore output arguments
181 //
182 A = A_org;
183
184 //
185 // Compare results
186 //
187 lauu2_native(A);
188
189 bool failed = false;
190 if (! isIdentical(A_generic, A, "A_generic", "_A")) {
191 std::cerr << "A_org = " << A_org << std::endl;
192 std::cerr << "CXXLAPACK: A_generic = " << A_generic << std::endl;
193 std::cerr << "F77LAPACK: A = " << A << std::endl;
194 failed = true;
195 }
196
197 if (failed) {
198 ASSERT(0);
199 }
200
201 # endif
202 }
203
204 //-- forwarding ----------------------------------------------------------------
205 template <typename MA>
206 void
207 lauu2(MA &&A)
208 {
209 lauu2(A);
210 }
211
212 } } // namespace lapack, flens
213
214 #endif // FLENS_LAPACK_GESV_LAUU2_TCC
2 * Copyright (c) 2011, Michael Lehn
3 *
4 * All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 *
10 * 1) Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2) Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in
14 * the documentation and/or other materials provided with the
15 * distribution.
16 * 3) Neither the name of the FLENS development group nor the names of
17 * its contributors may be used to endorse or promote products derived
18 * from this software without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
23 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
24 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
25 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
26 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
27 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
28 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
29 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
30 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
31 */
32
33 /* Based on
34 *
35 SUBROUTINE DLAUU2( UPLO, N, A, LDA, INFO )
36 *
37 * -- LAPACK auxiliary routine (version 3.3.1) --
38 * -- LAPACK is a software package provided by Univ. of Tennessee, --
39 * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
40 * -- April 2011
41 */
42
43 #ifndef FLENS_LAPACK_GESV_LAUU2_TCC
44 #define FLENS_LAPACK_GESV_LAUU2_TCC 1
45
46 #include <algorithm>
47 #include <flens/blas/blas.h>
48 #include <flens/lapack/lapack.h>
49
50 #include <flens/lapack/interface/include/f77lapack.h>
51
52 namespace flens { namespace lapack {
53
54 //== generic lapack implementation =============================================
55
56 template <typename MA>
57 void
58 lauu2_generic(TrMatrix<MA> &A)
59 {
60 typedef typename TrMatrix<MA>::ElementType ElementType;
61 typedef typename TrMatrix<MA>::IndexType IndexType;
62
63 const ElementType One(1);
64 const IndexType n = A.dim();
65 const Underscore<IndexType> _;
66 //
67 // Quick return if possible
68 //
69 if (n==0) {
70 return;
71 }
72
73 if (A.upLo()==Upper) {
74 //
75 // Compute the product U * U**T.
76 //
77 for (IndexType i=1; i<=n; ++i) {
78 const ElementType a22 = A(i,i);
79 if (i<n) {
80 A(i,i) = A(i,_(i,n)) * A(i,_(i,n));
81
82 const auto range1 = _(1,i-1);
83 const auto range2 = i;
84 const auto range3 = _(i+1,n);
85
86 auto a12 = A(range1,range2);
87 const auto A13 = A(range1,range3);
88 const auto a23 = A(range2,range3);
89 blas::mv(NoTrans, One, A13, a23, a22, a12);
90 } else {
91 A(_,n) *= a22;
92 }
93 }
94 } else {
95 //
96 // Compute the product L**T * L.
97 //
98 for (IndexType i=1; i<=n; ++i) {
99 const ElementType a22 = A(i,i);
100 if (i<n) {
101 A(i,i) = A(_(i,n),i) * A(_(i,n),i);
102
103 const auto range1 = _(1,i-1);
104 const auto range2 = i;
105 const auto range3 = _(i+1,n);
106
107 auto a21 = A(range2,range1);
108 const auto A31 = A(range3,range1);
109 const auto a32 = A(range3,range2);
110
111 blas::mv(Trans, One, A31, a32, a22, a21);
112 } else {
113 A(n,_) *= a22;
114 }
115 }
116 }
117 }
118
119 //== interface for native lapack ===============================================
120
121 #ifdef CHECK_CXXLAPACK
122
123 template <typename MA>
124 void
125 lauu2_native(TrMatrix<MA> &A)
126 {
127 typedef typename TrMatrix<MA>::ElementType ElementType;
128
129 const char UPLO(A.upLo());
130 const INTEGER N = A.dim();
131 const INTEGER LDA = A.leadingDimension();
132 INTEGER INFO;
133
134 if (IsSame<ElementType,double>::value) {
135 LAPACK_DECL(dlauu2)(&UPLO,
136 &N,
137 A.data(),
138 &LDA,
139 &INFO);
140 } else {
141 ASSERT(0);
142 }
143 ASSERT(INFO==0);
144 }
145
146 #endif // CHECK_CXXLAPACK
147
148 //== public interface ==========================================================
149
150 template <typename MA>
151 void
152 lauu2(TrMatrix<MA> &A)
153 {
154 typedef typename TrMatrix<MA>::IndexType IndexType;
155
156 //
157 // Test the input parameters
158 //
159 ASSERT(A.firstRow()==1);
160 ASSERT(A.firstCol()==1);
161
162 # ifdef CHECK_CXXLAPACK
163 //
164 // Make copies of output arguments
165 //
166 typename TrMatrix<MA>::NoView A_org = A;
167 # endif
168
169 //
170 // Call implementation
171 //
172 lauu2_generic(A);
173
174 # ifdef CHECK_CXXLAPACK
175 //
176 // Make copies of generic results
177 //
178 typename TrMatrix<MA>::NoView A_generic = A;
179 //
180 // Restore output arguments
181 //
182 A = A_org;
183
184 //
185 // Compare results
186 //
187 lauu2_native(A);
188
189 bool failed = false;
190 if (! isIdentical(A_generic, A, "A_generic", "_A")) {
191 std::cerr << "A_org = " << A_org << std::endl;
192 std::cerr << "CXXLAPACK: A_generic = " << A_generic << std::endl;
193 std::cerr << "F77LAPACK: A = " << A << std::endl;
194 failed = true;
195 }
196
197 if (failed) {
198 ASSERT(0);
199 }
200
201 # endif
202 }
203
204 //-- forwarding ----------------------------------------------------------------
205 template <typename MA>
206 void
207 lauu2(MA &&A)
208 {
209 lauu2(A);
210 }
211
212 } } // namespace lapack, flens
213
214 #endif // FLENS_LAPACK_GESV_LAUU2_TCC