1 // Copyright 2010-2021 Google LLC
2 // Licensed under the Apache License, Version 2.0 (the "License");
3 // you may not use this file except in compliance with the License.
4 // You may obtain a copy of the License at
5 //
6 // http://www.apache.org/licenses/LICENSE-2.0
7 //
8 // Unless required by applicable law or agreed to in writing, software
9 // distributed under the License is distributed on an "AS IS" BASIS,
10 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
11 // See the License for the specific language governing permissions and
12 // limitations under the License.
13
14 #ifndef OR_TOOLS_SAT_SYNCHRONIZATION_H_
15 #define OR_TOOLS_SAT_SYNCHRONIZATION_H_
16
17 #include <cstdint>
18 #include <deque>
19 #include <limits>
20 #include <string>
21 #include <vector>
22
23 #include "absl/random/bit_gen_ref.h"
24 #include "absl/random/random.h"
25 #include "absl/synchronization/mutex.h"
26 #include "ortools/base/integral_types.h"
27 #include "ortools/base/logging.h"
28 #include "ortools/base/stl_util.h"
29 #include "ortools/sat/cp_model.pb.h"
30 #include "ortools/sat/integer.h"
31 #include "ortools/sat/model.h"
32 #include "ortools/sat/sat_base.h"
33 #include "ortools/sat/sat_parameters.pb.h"
34 #include "ortools/sat/util.h"
35 #include "ortools/util/bitset.h"
36 #include "ortools/util/logging.h"
37 #include "ortools/util/time_limit.h"
38
39 namespace operations_research {
40 namespace sat {
41
42 // Thread-safe. Keeps a set of n unique best solution found so far.
43 //
44 // TODO(user): Maybe add some criteria to only keep solution with an objective
45 // really close to the best solution.
46 template <typename ValueType>
47 class SharedSolutionRepository {
48 public:
SharedSolutionRepository(int num_solutions_to_keep)49 explicit SharedSolutionRepository(int num_solutions_to_keep)
50 : num_solutions_to_keep_(num_solutions_to_keep) {
51 CHECK_GE(num_solutions_to_keep_, 0);
52 }
53
54 // The solution format used by this class.
55 struct Solution {
56 // Solution with lower "rank" will be preferred
57 //
58 // TODO(user): Some LNS code assume that for the SharedSolutionRepository
59 // this rank is actually the unscaled internal minimization objective.
60 // Remove this assumptions by simply recomputing this value since it is not
61 // too costly to do so.
62 int64_t rank = 0;
63
64 std::vector<ValueType> variable_values;
65
66 // Number of time this was returned by GetRandomBiasedSolution(). We use
67 // this information during the selection process.
68 //
69 // Should be private: only SharedSolutionRepository should modify this.
70 mutable int num_selected = 0;
71
72 bool operator==(const Solution& other) const {
73 return rank == other.rank && variable_values == other.variable_values;
74 }
75 bool operator<(const Solution& other) const {
76 if (rank != other.rank) {
77 return rank < other.rank;
78 }
79 return variable_values < other.variable_values;
80 }
81 };
82
83 // Returns the number of current solution in the pool. This will never
84 // decrease.
85 int NumSolutions() const;
86
87 // Returns the solution #i where i must be smaller than NumSolutions().
88 Solution GetSolution(int index) const;
89
90 // Returns the variable value of variable 'var_index' from solution
91 // 'solution_index' where solution_index must be smaller than NumSolutions()
92 // and 'var_index' must be smaller than number of variables.
93 ValueType GetVariableValueInSolution(int var_index, int solution_index) const;
94
95 // Returns a random solution biased towards good solutions.
96 Solution GetRandomBiasedSolution(absl::BitGenRef random) const;
97
98 // Add a new solution. Note that it will not be added to the pool of solution
99 // right away. One must call Synchronize for this to happen.
100 //
101 // Works in O(num_solutions_to_keep_).
102 void Add(const Solution& solution);
103
104 // Updates the current pool of solution with the one recently added. Note that
105 // we use a stable ordering of solutions, so the final pool will be
106 // independent on the order of the calls to AddSolution() provided that the
107 // set of added solutions is the same.
108 //
109 // Works in O(num_solutions_to_keep_).
110 void Synchronize();
111
112 protected:
113 // Helper method for adding the solutions once the mutex is acquired.
114 void AddInternal(const Solution& solution)
115 ABSL_EXCLUSIVE_LOCKS_REQUIRED(mutex_);
116
117 const int num_solutions_to_keep_;
118 mutable absl::Mutex mutex_;
119 int64_t num_synchronization_ ABSL_GUARDED_BY(mutex_) = 0;
120
121 // Our two solutions pools, the current one and the new one that will be
122 // merged into the current one on each Synchronize() calls.
123 mutable std::vector<int> tmp_indices_ ABSL_GUARDED_BY(mutex_);
124 std::vector<Solution> solutions_ ABSL_GUARDED_BY(mutex_);
125 std::vector<Solution> new_solutions_ ABSL_GUARDED_BY(mutex_);
126 };
127
128 // This is currently only used to store feasible solution from our 'relaxation'
129 // LNS generators which in turn are used to generate some RINS neighborhood.
130 class SharedRelaxationSolutionRepository
131 : public SharedSolutionRepository<int64_t> {
132 public:
SharedRelaxationSolutionRepository(int num_solutions_to_keep)133 explicit SharedRelaxationSolutionRepository(int num_solutions_to_keep)
134 : SharedSolutionRepository<int64_t>(num_solutions_to_keep) {}
135
136 void NewRelaxationSolution(const CpSolverResponse& response);
137 };
138
139 class SharedLPSolutionRepository : public SharedSolutionRepository<double> {
140 public:
SharedLPSolutionRepository(int num_solutions_to_keep)141 explicit SharedLPSolutionRepository(int num_solutions_to_keep)
142 : SharedSolutionRepository<double>(num_solutions_to_keep) {}
143
144 void NewLPSolution(std::vector<double> lp_solution);
145 };
146
147 // Set of partly filled solutions. They are meant to be finished by some lns
148 // worker.
149 //
150 // The solutions are stored as a vector of doubles. The value at index i
151 // represents the solution value of model variable indexed i. Note that some
152 // values can be infinity which should be interpreted as 'unknown' solution
153 // value for that variable. These solutions can not necessarily be completed to
154 // complete feasible solutions.
155 class SharedIncompleteSolutionManager {
156 public:
157 bool HasNewSolution() const;
158 std::vector<double> GetNewSolution();
159
160 void AddNewSolution(const std::vector<double>& lp_solution);
161
162 private:
163 // New solutions are added and removed from the back.
164 std::vector<std::vector<double>> solutions_;
165 mutable absl::Mutex mutex_;
166 };
167
168 // Manages the global best response kept by the solver. This class is
169 // responsible for logging the progress of the solutions and bounds as they are
170 // found.
171 //
172 // All functions are thread-safe except if specified otherwise.
173 class SharedResponseManager {
174 public:
175 explicit SharedResponseManager(Model* model);
176
177 // Loads the initial objective bounds and keep a reference to the objective to
178 // properly display the scaled bounds. This is optional if the model has no
179 // objective.
180 //
181 // This function is not thread safe.
182 void InitializeObjective(const CpModelProto& cp_model);
183
184 // Reports OPTIMAL and stop the search if any gap limit are specified and
185 // crossed. By default, we only stop when we have the true optimal, which is
186 // well defined since we are solving our pure integer problem exactly.
187 void SetGapLimitsFromParameters(const SatParameters& parameters);
188
189 // Returns the current solver response. That is the best known response at the
190 // time of the call with the best feasible solution and objective bounds.
191 //
192 // Note that the solver statistics correspond to the last time a better
193 // solution was found or SetStatsFromModel() was called.
194 //
195 // If full response is true, we will do more postprocessing by calling all the
196 // AddFinalSolutionPostprocessor() postprocesors. Note that the response given
197 // to the AddSolutionCallback() will not call them.
198 CpSolverResponse GetResponse(bool full_response = true);
199
200 // These will be called in REVERSE order on any feasible solution returned
201 // to the user.
202 void AddSolutionPostprocessor(
203 std::function<void(std::vector<int64_t>*)> postprocessor);
204
205 // These "postprocessing" steps will be applied in REVERSE order of
206 // registration to all solution passed to the callbacks.
207 void AddResponsePostprocessor(
208 std::function<void(CpSolverResponse*)> postprocessor);
209
210 // These "postprocessing" steps will only be applied after the others to the
211 // solution returned by GetResponse().
212 void AddFinalResponsePostprocessor(
213 std::function<void(CpSolverResponse*)> postprocessor);
214
215 // Adds a callback that will be called on each new solution (for
216 // statisfiablity problem) or each improving new solution (for an optimization
217 // problem). Returns its id so it can be unregistered if needed.
218 //
219 // Note that adding a callback is not free since the solution will be
220 // postsolved before this is called.
221 //
222 // Note that currently the class is waiting for the callback to finish before
223 // accepting any new updates. That could be changed if needed.
224 int AddSolutionCallback(
225 std::function<void(const CpSolverResponse&)> callback);
226 void UnregisterCallback(int callback_id);
227
228 // The "inner" objective is the CpModelProto objective without scaling/offset.
229 // Note that these bound correspond to valid bound for the problem of finding
230 // a strictly better objective than the current one. Thus the lower bound is
231 // always a valid bound for the global problem, but the upper bound is NOT.
232 IntegerValue GetInnerObjectiveLowerBound();
233 IntegerValue GetInnerObjectiveUpperBound();
234
235 // These functions return the same as the non-synchronized() version but
236 // only the values at the last time Synchronize() was called.
237 void Synchronize();
238 IntegerValue SynchronizedInnerObjectiveLowerBound();
239 IntegerValue SynchronizedInnerObjectiveUpperBound();
240
241 // Returns the current best solution inner objective value or kInt64Max if
242 // there is no solution.
243 IntegerValue BestSolutionInnerObjectiveValue();
244
245 // Returns the integral of the log of the absolute gap over deterministic
246 // time. This is mainly used to compare how fast the gap closes on a
247 // particular instance. Or to evaluate how efficient our LNS code is improving
248 // solution.
249 //
250 // Note: The integral will start counting on the first UpdateGapIntegral()
251 // call, since before the difference is assumed to be zero.
252 //
253 // Important: To report a proper deterministic integral, we only update it
254 // on UpdateGapIntegral() which should be called in the main subsolver
255 // synchronization loop.
256 //
257 // Note(user): In the litterature, people use the relative gap to the optimal
258 // solution (or the best known one), but this is ill defined in many case
259 // (like if the optimal cost is zero), so I prefer this version.
260 double GapIntegral() const;
261 void UpdateGapIntegral();
262
263 // Sets this to true to have the "real" but non-deterministic primal integral.
264 // If this is true, then there is no need to manually call
265 // UpdateGapIntegral() but it is not an issue to do so.
266 void SetUpdateGapIntegralOnEachChange(bool set);
267
268 // Updates the inner objective bounds.
269 void UpdateInnerObjectiveBounds(const std::string& update_info,
270 IntegerValue lb, IntegerValue ub);
271
272 // Reads the new solution from the response and update our state. For an
273 // optimization problem, we only do something if the solution is strictly
274 // improving.
275 //
276 // TODO(user): Only the following fields from response are accessed here, we
277 // might want a tighter API:
278 // - solution_info
279 // - solution
280 void NewSolution(const CpSolverResponse& response, Model* model);
281
282 // Changes the solution to reflect the fact that the "improving" problem is
283 // infeasible. This means that if we have a solution, we have proven
284 // optimality, otherwise the global problem is infeasible.
285 //
286 // Note that this shouldn't be called before the solution is actually
287 // reported. We check for this case in NewSolution().
288 void NotifyThatImprovingProblemIsInfeasible(const std::string& worker_info);
289
290 // Adds to the shared response a subset of assumptions that are enough to
291 // make the problem infeasible.
292 void AddUnsatCore(const std::vector<int>& core);
293
294 // Sets the statistics in the response to the one of the solver inside the
295 // given in-memory model. This does nothing if the model is nullptr.
296 //
297 // TODO(user): Also support merging statistics together.
298 void SetStatsFromModel(Model* model);
299
300 // Returns true if we found the optimal solution or the problem was proven
301 // infeasible. Note that if the gap limit is reached, we will also report
302 // OPTIMAL and consider the problem solved.
303 bool ProblemIsSolved() const;
304
305 // Returns the underlying solution repository where we keep a set of best
306 // solutions.
SolutionsRepository()307 const SharedSolutionRepository<int64_t>& SolutionsRepository() const {
308 return solutions_;
309 }
MutableSolutionsRepository()310 SharedSolutionRepository<int64_t>* MutableSolutionsRepository() {
311 return &solutions_;
312 }
313
314 // This should be called after the model is loaded. It will read the file
315 // specified by --cp_model_load_debug_solution and properly fill the
316 // model->Get<DebugSolution>() vector.
317 //
318 // TODO(user): Note that for now, only the IntegerVariable value are loaded,
319 // not the value of the pure Booleans variables.
320 void LoadDebugSolution(Model*);
321
322 // Debug only. Set dump prefix for solutions written to file.
set_dump_prefix(const std::string & dump_prefix)323 void set_dump_prefix(const std::string& dump_prefix) {
324 dump_prefix_ = dump_prefix;
325 }
326
327 // Display improvement stats.
328 void DisplayImprovementStatistics();
329
330 void LogMessage(std::string message);
331
332 // This is here for the few codepath that needs to modify the returned
333 // response directly. Note that this do not work in parallel.
334 //
335 // TODO(user): This can probably be removed.
MutableResponse()336 CpSolverResponse* MutableResponse() {
337 absl::MutexLock mutex_lock(&mutex_);
338 return &best_response_;
339 }
340
341 private:
342 void TestGapLimitsIfNeeded() ABSL_EXCLUSIVE_LOCKS_REQUIRED(mutex_);
343 void FillObjectiveValuesInBestResponse()
344 ABSL_EXCLUSIVE_LOCKS_REQUIRED(mutex_);
345 void SetStatsFromModelInternal(Model* model)
346 ABSL_EXCLUSIVE_LOCKS_REQUIRED(mutex_);
347 void UpdateGapIntegralInternal() ABSL_EXCLUSIVE_LOCKS_REQUIRED(mutex_);
348
349 void RegisterSolutionFound(const std::string& improvement_info)
350 ABSL_EXCLUSIVE_LOCKS_REQUIRED(mutex_);
351 void RegisterObjectiveBoundImprovement(const std::string& improvement_info)
352 ABSL_EXCLUSIVE_LOCKS_REQUIRED(mutex_);
353
354 // Generates a response for callbacks and GetResponse().
355 CpSolverResponse GetResponseInternal() ABSL_EXCLUSIVE_LOCKS_REQUIRED(mutex_);
356
357 const SatParameters& parameters_;
358 const WallTimer& wall_timer_;
359 ModelSharedTimeLimit* shared_time_limit_;
360 CpObjectiveProto const* objective_or_null_ = nullptr;
361
362 mutable absl::Mutex mutex_;
363
364 // Gap limits.
365 double absolute_gap_limit_ ABSL_GUARDED_BY(mutex_) = 0.0;
366 double relative_gap_limit_ ABSL_GUARDED_BY(mutex_) = 0.0;
367
368 CpSolverResponse best_response_ ABSL_GUARDED_BY(mutex_);
369 SharedSolutionRepository<int64_t> solutions_ ABSL_GUARDED_BY(mutex_);
370
371 int num_solutions_ ABSL_GUARDED_BY(mutex_) = 0;
372 int64_t inner_objective_lower_bound_ ABSL_GUARDED_BY(mutex_) =
373 std::numeric_limits<int64_t>::min();
374 int64_t inner_objective_upper_bound_ ABSL_GUARDED_BY(mutex_) =
375 std::numeric_limits<int64_t>::max();
376 int64_t best_solution_objective_value_ ABSL_GUARDED_BY(mutex_) =
377 std::numeric_limits<int64_t>::max();
378
379 IntegerValue synchronized_inner_objective_lower_bound_ ABSL_GUARDED_BY(
380 mutex_) = IntegerValue(std::numeric_limits<int64_t>::min());
381 IntegerValue synchronized_inner_objective_upper_bound_ ABSL_GUARDED_BY(
382 mutex_) = IntegerValue(std::numeric_limits<int64_t>::max());
383
384 bool update_integral_on_each_change_ ABSL_GUARDED_BY(mutex_) = false;
385 double gap_integral_ ABSL_GUARDED_BY(mutex_) = 0.0;
386 double last_absolute_gap_ ABSL_GUARDED_BY(mutex_) = 0.0;
387 double last_gap_integral_time_stamp_ ABSL_GUARDED_BY(mutex_) = 0.0;
388
389 int next_callback_id_ ABSL_GUARDED_BY(mutex_) = 0;
390 std::vector<std::pair<int, std::function<void(const CpSolverResponse&)>>>
391 callbacks_ ABSL_GUARDED_BY(mutex_);
392
393 std::vector<std::function<void(std::vector<int64_t>*)>>
394 solution_postprocessors_ ABSL_GUARDED_BY(mutex_);
395 std::vector<std::function<void(CpSolverResponse*)>> postprocessors_
396 ABSL_GUARDED_BY(mutex_);
397 std::vector<std::function<void(CpSolverResponse*)>> final_postprocessors_
398 ABSL_GUARDED_BY(mutex_);
399
400 // Dump prefix.
401 std::string dump_prefix_;
402
403 // Used for statistics of the improvements found by workers.
404 std::map<std::string, int> primal_improvements_count_ ABSL_GUARDED_BY(mutex_);
405 std::map<std::string, int> dual_improvements_count_ ABSL_GUARDED_BY(mutex_);
406
407 SolverLogger* logger_;
408 };
409
410 // This class manages a pool of lower and upper bounds on a set of variables in
411 // a parallel context.
412 class SharedBoundsManager {
413 public:
414 explicit SharedBoundsManager(const CpModelProto& model_proto);
415
416 // Reports a set of locally improved variable bounds to the shared bounds
417 // manager. The manager will compare these bounds changes against its
418 // global state, and incorporate the improving ones.
419 void ReportPotentialNewBounds(const CpModelProto& model_proto,
420 const std::string& worker_name,
421 const std::vector<int>& variables,
422 const std::vector<int64_t>& new_lower_bounds,
423 const std::vector<int64_t>& new_upper_bounds);
424
425 // If we solved a small independent component of the full problem, then we can
426 // in most situation fix the solution on this subspace.
427 //
428 // Note that because there can be more than one optimal solution on an
429 // independent subproblem, it is important to do that in a locked fashion, and
430 // reject future incompatible fixing.
431 void FixVariablesFromPartialSolution(
432 const std::vector<int64_t>& solution,
433 const std::vector<int>& variables_to_fix);
434
435 // Returns a new id to be used in GetChangedBounds(). This is just an ever
436 // increasing sequence starting from zero. Note that the class is not designed
437 // to have too many of these.
438 int RegisterNewId();
439
440 // When called, returns the set of bounds improvements since
441 // the last time this method was called with the same id.
442 void GetChangedBounds(int id, std::vector<int>* variables,
443 std::vector<int64_t>* new_lower_bounds,
444 std::vector<int64_t>* new_upper_bounds);
445
446 // Publishes any new bounds so that GetChangedBounds() will reflect the latest
447 // state.
448 void Synchronize();
449
450 private:
451 const int num_variables_;
452 const CpModelProto& model_proto_;
453
454 absl::Mutex mutex_;
455
456 // These are always up to date.
457 std::vector<int64_t> lower_bounds_ ABSL_GUARDED_BY(mutex_);
458 std::vector<int64_t> upper_bounds_ ABSL_GUARDED_BY(mutex_);
459 SparseBitset<int64_t> changed_variables_since_last_synchronize_
460 ABSL_GUARDED_BY(mutex_);
461
462 // These are only updated on Synchronize().
463 std::vector<int64_t> synchronized_lower_bounds_ ABSL_GUARDED_BY(mutex_);
464 std::vector<int64_t> synchronized_upper_bounds_ ABSL_GUARDED_BY(mutex_);
465 std::deque<SparseBitset<int64_t>> id_to_changed_variables_
466 ABSL_GUARDED_BY(mutex_);
467 };
468
469 template <typename ValueType>
NumSolutions()470 int SharedSolutionRepository<ValueType>::NumSolutions() const {
471 absl::MutexLock mutex_lock(&mutex_);
472 return solutions_.size();
473 }
474
475 template <typename ValueType>
476 typename SharedSolutionRepository<ValueType>::Solution
GetSolution(int i)477 SharedSolutionRepository<ValueType>::GetSolution(int i) const {
478 absl::MutexLock mutex_lock(&mutex_);
479 return solutions_[i];
480 }
481
482 template <typename ValueType>
GetVariableValueInSolution(int var_index,int solution_index)483 ValueType SharedSolutionRepository<ValueType>::GetVariableValueInSolution(
484 int var_index, int solution_index) const {
485 absl::MutexLock mutex_lock(&mutex_);
486 return solutions_[solution_index].variable_values[var_index];
487 }
488
489 // TODO(user): Experiments on the best distribution.
490 template <typename ValueType>
491 typename SharedSolutionRepository<ValueType>::Solution
GetRandomBiasedSolution(absl::BitGenRef random)492 SharedSolutionRepository<ValueType>::GetRandomBiasedSolution(
493 absl::BitGenRef random) const {
494 absl::MutexLock mutex_lock(&mutex_);
495 const int64_t best_rank = solutions_[0].rank;
496
497 // As long as we have solution with the best objective that haven't been
498 // explored too much, we select one uniformly. Otherwise, we select a solution
499 // from the pool uniformly.
500 //
501 // Note(user): Because of the increase of num_selected, this is dependent on
502 // the order of call. It should be fine for "determinism" because we do
503 // generate the task of a batch always in the same order.
504 const int kExplorationThreshold = 100;
505
506 // Select all the best solution with a low enough selection count.
507 tmp_indices_.clear();
508 for (int i = 0; i < solutions_.size(); ++i) {
509 const auto& solution = solutions_[i];
510 if (solution.rank == best_rank &&
511 solution.num_selected <= kExplorationThreshold) {
512 tmp_indices_.push_back(i);
513 }
514 }
515
516 int index = 0;
517 if (tmp_indices_.empty()) {
518 index = absl::Uniform<int>(random, 0, solutions_.size());
519 } else {
520 index = tmp_indices_[absl::Uniform<int>(random, 0, tmp_indices_.size())];
521 }
522 solutions_[index].num_selected++;
523 return solutions_[index];
524 }
525
526 template <typename ValueType>
Add(const Solution & solution)527 void SharedSolutionRepository<ValueType>::Add(const Solution& solution) {
528 if (num_solutions_to_keep_ == 0) return;
529 absl::MutexLock mutex_lock(&mutex_);
530 AddInternal(solution);
531 }
532
533 template <typename ValueType>
AddInternal(const Solution & solution)534 void SharedSolutionRepository<ValueType>::AddInternal(
535 const Solution& solution) {
536 int worse_solution_index = 0;
537 for (int i = 0; i < new_solutions_.size(); ++i) {
538 // Do not add identical solution.
539 if (new_solutions_[i] == solution) return;
540 if (new_solutions_[worse_solution_index] < new_solutions_[i]) {
541 worse_solution_index = i;
542 }
543 }
544 if (new_solutions_.size() < num_solutions_to_keep_) {
545 new_solutions_.push_back(solution);
546 } else if (solution < new_solutions_[worse_solution_index]) {
547 new_solutions_[worse_solution_index] = solution;
548 }
549 }
550
551 template <typename ValueType>
Synchronize()552 void SharedSolutionRepository<ValueType>::Synchronize() {
553 absl::MutexLock mutex_lock(&mutex_);
554 if (new_solutions_.empty()) return;
555
556 solutions_.insert(solutions_.end(), new_solutions_.begin(),
557 new_solutions_.end());
558 new_solutions_.clear();
559
560 // We use a stable sort to keep the num_selected count for the already
561 // existing solutions.
562 //
563 // TODO(user): Introduce a notion of orthogonality to diversify the pool?
564 gtl::STLStableSortAndRemoveDuplicates(&solutions_);
565 if (solutions_.size() > num_solutions_to_keep_) {
566 solutions_.resize(num_solutions_to_keep_);
567 }
568
569 if (!solutions_.empty()) {
570 VLOG(2) << "Solution pool update:"
571 << " num_solutions=" << solutions_.size()
572 << " min_rank=" << solutions_[0].rank
573 << " max_rank=" << solutions_.back().rank;
574 }
575
576 num_synchronization_++;
577 }
578
579 } // namespace sat
580 } // namespace operations_research
581
582 #endif // OR_TOOLS_SAT_SYNCHRONIZATION_H_
583