Dune Core Modules (2.9.1)

cholmod.hh
1// SPDX-FileCopyrightText: Copyright (C) DUNE Project contributors, see file LICENSE.md in module root
2// SPDX-License-Identifier: LicenseRef-GPL-2.0-only-with-DUNE-exception
3#pragma once
4
5#if HAVE_SUITESPARSE_CHOLMOD
6
10#include <dune/istl/bvector.hh>
11#include<dune/istl/solver.hh>
12#include <dune/istl/solverfactory.hh>
13#include <dune/istl/foreach.hh>
14
15#include <vector>
16#include <memory>
17
18#include <cholmod.h>
19
20namespace Dune {
21
22namespace Impl{
23
32 struct NoIgnore
33 {
34 const NoIgnore& operator[](std::size_t) const { return *this; }
35 explicit operator bool() const { return false; }
36 static constexpr std::size_t size() { return 0; }
37
38 };
39
40
41 template<class BlockedVector, class FlatVector>
42 void copyToFlatVector(const BlockedVector& blockedVector, FlatVector& flatVector)
43 {
44 // traverse the vector once just to compute the size
45 std::size_t len = flatVectorForEach(blockedVector, [&](auto&&, auto...){});
46 flatVector.resize(len);
47
48 flatVectorForEach(blockedVector, [&](auto&& entry, auto offset){
49 flatVector[offset] = entry;
50 });
51 }
52
53 // special (dummy) case for NoIgnore
54 template<class FlatVector>
55 void copyToFlatVector(const NoIgnore&, FlatVector&)
56 {
57 // just do nothing
58 return;
59 }
60
61 template<class FlatVector, class BlockedVector>
62 void copyToBlockedVector(const FlatVector& flatVector, BlockedVector& blockedVector)
63 {
64 flatVectorForEach(blockedVector, [&](auto& entry, auto offset){
65 entry = flatVector[offset];
66 });
67 }
68
69
70} //namespace Impl
71
76template<class Vector>
77class Cholmod : public InverseOperator<Vector, Vector>
78{
79public:
80
86 Cholmod()
87 {
88 cholmod_start(&c_);
89 }
90
96 ~Cholmod()
97 {
98 if (L_)
99 cholmod_free_factor(&L_, &c_);
100 cholmod_finish(&c_);
101 }
102
103 // forbid copying to avoid freeing memory twice
104 Cholmod(const Cholmod&) = delete;
105 Cholmod& operator=(const Cholmod&) = delete;
106
107
110 void apply (Vector& x, Vector& b, [[maybe_unused]] double reduction, InverseOperatorResult& res)
111 {
112 apply(x,b,res);
113 }
114
120 void apply(Vector& x, Vector& b, InverseOperatorResult& res)
121 {
122 // do nothing if N=0
123 if ( nIsZero_ )
124 {
125 return;
126 }
127
128 if (x.size() != b.size())
129 DUNE_THROW(Exception, "Error in apply(): sizes of x and b do not match!");
130
131 // cast to double array
132 auto b2 = std::make_unique<double[]>(L_->n);
133 auto x2 = std::make_unique<double[]>(L_->n);
134
135 // copy to cholmod
136 auto bp = b2.get();
137
138 flatVectorForEach(b, [&](auto&& entry, auto&& flatIndex){
139 if ( subIndices_.empty() )
140 bp[ flatIndex ] = entry;
141 else
142 if( subIndices_[ flatIndex ] != std::numeric_limits<std::size_t>::max() )
143 bp[ subIndices_[ flatIndex ] ] = entry;
144 });
145
146 // create a cholmod dense object
147 auto b3 = make_cholmod_dense(cholmod_allocate_dense(L_->n, 1, L_->n, CHOLMOD_REAL, &c_), &c_);
148 // cast because void-ptr
149 auto b4 = static_cast<double*>(b3->x);
150 std::copy(b2.get(), b2.get() + L_->n, b4);
151
152 // solve for a cholmod x object
153 auto x3 = make_cholmod_dense(cholmod_solve(CHOLMOD_A, L_, b3.get(), &c_), &c_);
154 // cast because void-ptr
155 auto xp = static_cast<double*>(x3->x);
156
157 // copy into x
158 flatVectorForEach(x, [&](auto&& entry, auto&& flatIndex){
159 if ( subIndices_.empty() )
160 entry = xp[ flatIndex ];
161 else
162 if( subIndices_[ flatIndex ] != std::numeric_limits<std::size_t>::max() )
163 entry = xp[ subIndices_[ flatIndex ] ];
164 });
165
166 // statistics for a direct solver
167 res.iterations = 1;
168 res.converged = true;
169 }
170
171
177 template<class Matrix>
178 void setMatrix(const Matrix& matrix)
179 {
180 const Impl::NoIgnore* noIgnore = nullptr;
181 setMatrix(matrix, noIgnore);
182 }
183
198 template<class Matrix, class Ignore>
199 void setMatrix(const Matrix& matrix, const Ignore* ignore)
200 {
201 // count the number of entries and diagonal entries
202 int nonZeros = 0;
203 int numberOfIgnoredDofs = 0;
204
205
206 auto [flatRows,flatCols] = flatMatrixForEach( matrix, [&](auto&& /*entry*/, auto&& flatRowIndex, auto&& flatColIndex){
207 if( flatRowIndex <= flatColIndex )
208 nonZeros++;
209 });
210
211 std::vector<bool> flatIgnore;
212
213 if ( ignore )
214 {
215 Impl::copyToFlatVector(*ignore,flatIgnore);
216 numberOfIgnoredDofs = std::count(flatIgnore.begin(),flatIgnore.end(),true);
217 }
218
219 // Total number of rows
220 int N = flatRows - numberOfIgnoredDofs;
221
222 nIsZero_ = (N <= 0);
223
224 if ( nIsZero_ )
225 {
226 return;
227 }
228
229 /*
230 * CHOLMOD uses compressed-column sparse matrices, but for symmetric
231 * matrices this is the same as the compressed-row sparse matrix used
232 * by DUNE. So we can just store Mᵀ instead of M (as M = Mᵀ).
233 */
234 const auto deleter = [c = &this->c_](auto* p) {
235 cholmod_free_sparse(&p, c);
236 };
237 auto M = std::unique_ptr<cholmod_sparse, decltype(deleter)>(
238 cholmod_allocate_sparse(N, // # rows
239 N, // # cols
240 nonZeros, // # of nonzeroes
241 1, // indices are sorted ( 1 = true)
242 1, // matrix is "packed" ( 1 = true)
243 -1, // stype of matrix ( -1 = consider the lower part only )
244 CHOLMOD_REAL, // xtype of matrix ( CHOLMOD_REAL = single array, no complex numbers)
245 &c_ // cholmod_common ptr
246 ), deleter);
247
248 // copy the data of BCRS matrix to Cholmod Sparse matrix
249 int* Ap = static_cast<int*>(M->p);
250 int* Ai = static_cast<int*>(M->i);
251 double* Ax = static_cast<double*>(M->x);
252
253
254 if ( ignore )
255 {
256 // init the mapping
257 subIndices_.resize(flatRows,std::numeric_limits<std::size_t>::max());
258
259 std::size_t subIndexCounter = 0;
260
261 for ( std::size_t i=0; i<flatRows; i++ )
262 {
263 if ( not flatIgnore[ i ] )
264 {
265 subIndices_[ i ] = subIndexCounter++;
266 }
267 }
268 }
269
270 // at first, we need to compute the row starts "Ap"
271 // therefore, we count all (not ignored) entries in each row and in the end we accumulate everything
272 flatMatrixForEach(matrix, [&](auto&& /*entry*/, auto&& flatRowIndex, auto&& flatColIndex){
273
274 // stop if ignored
275 if ( ignore and ( flatIgnore[flatRowIndex] or flatIgnore[flatColIndex] ) )
276 return;
277
278 // stop if in lower half
279 if ( flatRowIndex > flatColIndex )
280 return;
281
282 // ok, count the entry
283 auto idx = ignore ? subIndices_[flatRowIndex] : flatRowIndex;
284 Ap[idx+1]++;
285
286 });
287
288 // now accumulate
289 Ap[0] = 0;
290 for ( int i=0; i<N; i++ )
291 {
292 Ap[i+1] += Ap[i];
293 }
294
295 // we need a compressed row position counter
296 std::vector<std::size_t> rowPosition(N,0);
297
298 // now we can set the entries
299 flatMatrixForEach(matrix, [&](auto&& entry, auto&& flatRowIndex, auto&& flatColIndex){
300
301 // stop if ignored
302 if ( ignore and ( flatIgnore[flatRowIndex] or flatIgnore[flatColIndex] ) )
303 return;
304
305 // stop if in lower half
306 if ( flatRowIndex > flatColIndex )
307 return;
308
309 // ok, set the entry
310 auto rowIdx = ignore ? subIndices_[flatRowIndex] : flatRowIndex;
311 auto colIdx = ignore ? subIndices_[flatColIndex] : flatColIndex;
312 auto rowStart = Ap[rowIdx];
313 auto rowPos = rowPosition[rowIdx];
314 Ai[ rowStart + rowPos ] = colIdx;
315 Ax[ rowStart + rowPos ] = entry;
316 rowPosition[rowIdx]++;
317
318 });
319
320 // Now analyse the pattern and optimal row order
321 L_ = cholmod_analyze(M.get(), &c_);
322
323 // Do the factorization (this may take some time)
324 cholmod_factorize(M.get(), L_, &c_);
325 }
326
327 virtual SolverCategory::Category category() const
328 {
329 return SolverCategory::Category::sequential;
330 }
331
337 cholmod_common& cholmodCommonObject()
338 {
339 return c_;
340 }
341
347 cholmod_factor& cholmodFactor()
348 {
349 return *L_;
350 }
351
357 const cholmod_factor& cholmodFactor() const
358 {
359 return *L_;
360 }
361private:
362
363 // create a std::unique_ptr to a cholmod_dense object with a deleter
364 // that calls the appropriate cholmod cleanup routine
365 auto make_cholmod_dense(cholmod_dense* x, cholmod_common* c)
366 {
367 const auto deleter = [c](auto* p) {
368 cholmod_free_dense(&p, c);
369 };
370 return std::unique_ptr<cholmod_dense, decltype(deleter)>(x, deleter);
371 }
372
373 cholmod_common c_;
374 cholmod_factor* L_ = nullptr;
375
376 // indicator for a 0x0 problem (due to ignore dof's)
377 bool nIsZero_ = false;
378
379 // vector mapping all indices in flat order to the not ignored indices
380 std::vector<std::size_t> subIndices_;
381};
382
383 struct CholmodCreator{
384 template<class F> struct isValidBlock : std::false_type{};
385 template<int k> struct isValidBlock<FieldVector<double,k>> : std::true_type{};
386 template<int k> struct isValidBlock<FieldVector<float,k>> : std::true_type{};
387
388 template<class TL, typename M>
389 std::shared_ptr<Dune::InverseOperator<typename Dune::TypeListElement<1, TL>::type,
390 typename Dune::TypeListElement<2, TL>::type>>
391 operator()(TL /*tl*/, const M& mat, const Dune::ParameterTree& /*config*/,
392 std::enable_if_t<isValidBlock<typename Dune::TypeListElement<1, TL>::type::block_type>::value,int> = 0) const
393 {
394 using D = typename Dune::TypeListElement<1, TL>::type;
395 auto solver = std::make_shared<Dune::Cholmod<D>>();
396 solver->setMatrix(mat);
397 return solver;
398 }
399
400 // second version with SFINAE to validate the template parameters of Cholmod
401 template<typename TL, typename M>
402 std::shared_ptr<Dune::InverseOperator<typename Dune::TypeListElement<1, TL>::type,
403 typename Dune::TypeListElement<2, TL>::type>>
404 operator() (TL /*tl*/, const M& /*mat*/, const Dune::ParameterTree& /*config*/,
405 std::enable_if_t<!isValidBlock<typename Dune::TypeListElement<1, TL>::type::block_type>::value,int> = 0) const
406 {
407 DUNE_THROW(UnsupportedType, "Unsupported Type in Cholmod");
408 }
409 };
410 DUNE_REGISTER_DIRECT_SOLVER("cholmod", Dune::CholmodCreator());
411
412} /* namespace Dune */
413
414#endif // HAVE_SUITESPARSE_CHOLMOD
Implementation of the BCRSMatrix class.
This file implements a vector space as a tensor product of a given vector space. The number of compon...
Hierarchical structure of string parameters.
Definition: parametertree.hh:37
Implements a matrix constructed from a given type representing a field and compile-time given number ...
Implements a vector constructed from a given type representing a field and a compile-time given size.
#define DUNE_THROW(E, m)
Definition: exceptions.hh:218
auto max(ADLTag< 0 >, const V &v1, const V &v2)
implements binary Simd::max()
Definition: defaults.hh:81
Dune namespace.
Definition: alignedallocator.hh:13
std::pair< std::size_t, std::size_t > flatMatrixForEach(Matrix &&matrix, F &&f, std::size_t rowOffset=0, std::size_t colOffset=0)
Traverse a blocked matrix and call a functor at each scalar entry.
Definition: foreach.hh:132
std::size_t flatVectorForEach(Vector &&vector, F &&f, std::size_t offset=0)
Traverse a blocked vector and call a functor at each scalar entry.
Definition: foreach.hh:95
Define general, extensible interface for inverse operators.
Category
Definition: solvercategory.hh:23
Creative Commons License   |  Legal Statements / Impressum  |  Hosted by TU Dresden  |  generated with Hugo v0.111.3 (Nov 21, 23:30, 2024)