-/* Copyright (c) 2019-2020. The SimGrid Team. All rights reserved. */
+/* Copyright (c) 2019-2023. The SimGrid Team. All rights reserved. */
/* This program is free software; you can redistribute it and/or modify it
* under the terms of the license (GNU LGPL) which comes with this package. */
-#include "xbt/random.hpp"
#include "xbt/asserts.h"
+#include <fstream>
+#include <iostream>
#include <limits>
-#include <random>
+#include <memory>
+#include <string>
+#include <xbt/log.hpp>
+#include <xbt/random.hpp>
-namespace simgrid {
-namespace xbt {
-namespace random {
-enum xbt_random_implem { XBT_RNG_xbt, XBT_RNG_std };
-static xbt_random_implem rng_implem = XBT_RNG_xbt;
+XBT_LOG_EXTERNAL_CATEGORY(xbt);
+XBT_LOG_NEW_DEFAULT_SUBCATEGORY(xbt_random, xbt, "Random");
-static std::mt19937 mt19937_gen;
+namespace simgrid::xbt::random {
-void set_implem_xbt()
+bool Random::read_state(const std::string& filename)
{
- rng_implem = XBT_RNG_xbt;
+ std::ifstream file(filename);
+ file >> mt19937_gen;
+ file.close();
+ if (file.fail())
+ XBT_WARN("Could not save the RNG state to file %s.", filename.c_str());
+ return not file.fail();
}
-void set_implem_std()
+
+bool Random::write_state(const std::string& filename) const
{
- rng_implem = XBT_RNG_std;
+ std::ofstream file(filename);
+ file << mt19937_gen;
+ file.close();
+ if (file.fail())
+ XBT_WARN("Could not read the RNG state from file %s.", filename.c_str());
+ return not file.fail();
}
-void set_mersenne_seed(int seed)
+
+int StdRandom::uniform_int(int min, int max)
{
- mt19937_gen.seed(seed);
+ std::uniform_int_distribution dist(min, max);
+ return dist(mt19937_gen);
}
-int uniform_int(int min, int max)
+double StdRandom::uniform_real(double min, double max)
+{
+ std::uniform_real_distribution dist(min, max);
+ return dist(mt19937_gen);
+}
+
+double StdRandom::exponential(double lambda)
{
- if (rng_implem == XBT_RNG_std) {
- std::uniform_int_distribution<> dist(min, max);
- return dist(mt19937_gen);
- }
+ std::exponential_distribution dist(lambda);
+ return dist(mt19937_gen);
+}
- unsigned long range = max - min + 1;
+double StdRandom::normal(double mean, double sd)
+{
+ std::normal_distribution dist(mean, sd);
+ return dist(mt19937_gen);
+}
+
+int XbtRandom::uniform_int(int min, int max)
+{
+ // The casts to unsigned are here to ensure that the value of range is correctly calculated, even when greater than
+ // INT_MAX. See the corresponding unit tests for examples.
+ unsigned long range = static_cast<unsigned>(max) - static_cast<unsigned>(min);
xbt_assert(min <= max,
"The minimum value for the uniform integer distribution must not be greater than the maximum value");
- xbt_assert(range > 0, "Overflow in the uniform integer distribution, please use a smaller range.");
+ xbt_assert(range <= decltype(mt19937_gen)::max(),
+ "Overflow in the uniform integer distribution, please use a smaller range.");
+ if (range == decltype(mt19937_gen)::max())
+ return static_cast<int>(mt19937_gen() + min);
+
+ ++range;
+ unsigned long limit = decltype(mt19937_gen)::max() - decltype(mt19937_gen)::max() % range;
unsigned long value;
do {
value = mt19937_gen();
- } while (value >= decltype(mt19937_gen)::max() - decltype(mt19937_gen)::max() % range);
- return value % range + min;
+ } while (value >= limit);
+ return static_cast<int>(value % range + min);
}
-double uniform_real(double min, double max)
+double XbtRandom::uniform_real(double min, double max)
{
- if (rng_implem == XBT_RNG_std) {
- std::uniform_real_distribution<> dist(min, max);
- return dist(mt19937_gen);
- }
-
// This reuses Boost's uniform real distribution ideas
constexpr unsigned long divisor = decltype(mt19937_gen)::max() - decltype(mt19937_gen)::min();
unsigned long numerator;
do {
numerator = mt19937_gen() - decltype(mt19937_gen)::min();
} while (numerator == divisor);
- return min + (max - min) * numerator / divisor;
+ return min + (max - min) * static_cast<double>(numerator) / divisor;
}
-double exponential(double lambda)
+double XbtRandom::exponential(double lambda)
{
- if (rng_implem == XBT_RNG_std) {
- std::exponential_distribution<> dist(lambda);
- return dist(mt19937_gen);
- }
-
return -1.0 / lambda * log(uniform_real(0.0, 1.0));
}
-double normal(double mean, double sd)
+double XbtRandom::normal(double mean, double sd)
{
- if (rng_implem == XBT_RNG_std) {
- std::normal_distribution<> dist(mean, sd);
- return dist(mt19937_gen);
- }
-
double u1;
do {
u1 = uniform_real(0.0, 1.0);
return z0 * sd + mean;
}
-} // namespace random
-} // namespace xbt
-} // namespace simgrid
+static std::unique_ptr<Random> default_random = std::make_unique<XbtRandom>();
+
+void set_implem_xbt()
+{
+ default_random = std::make_unique<XbtRandom>();
+}
+void set_implem_std()
+{
+ default_random = std::make_unique<StdRandom>();
+}
+
+void set_mersenne_seed(int seed)
+{
+ default_random->set_seed(seed);
+}
+
+bool read_mersenne_state(const std::string& filename)
+{
+ return default_random->read_state(filename);
+}
+
+bool write_mersenne_state(const std::string& filename)
+{
+ return default_random->write_state(filename);
+}
+
+int uniform_int(int min, int max)
+{
+ return default_random->uniform_int(min, max);
+}
+
+double uniform_real(double min, double max)
+{
+ return default_random->uniform_real(min, max);
+}
+
+double exponential(double lambda)
+{
+ return default_random->exponential(lambda);
+}
+
+double normal(double mean, double sd)
+{
+ return default_random->normal(mean, sd);
+}
+
+} // namespace simgrid::xbt::random