使用Boost.Random从种子生成多精度整数

Question

I am attempting to use the Boost multiprecision libraries for C++ to generate large random numbers. I have been unable to create a generator which is seeded by the time or another random number, so my generator produces the same numbers on every run. How do I seed the generator with a changing value to produce different values on each run? Here is the code which works but produces the same values on each run:

 using namespace boost::multiprecision;
 using namespace boost::random;

 typedef independent_bits_engine<mt19937, 512, mpz_int> generator_type;
 generator_type gen;

 std::cout << gen() << "\n\n";

I have seeded the std mersenne twister before with success:

std::random_device rd;
std::mt19937 gen(rd());
std::uniform_int_distribution<> dis(1, 410);
std::cout << dis(gen);

but I am not sure how to seed the multiprecision mt. I get an error if I try to attach any arguments to the generator_type declaration.

Answer 1

Just be sure to include the right headers in the right order. Make sure boost/multiprecision/random.hpp is included before eg boost/random.hpp . ^[1]

Here's a working sample:

Live On Coliru

#include <boost/multiprecision/gmp.hpp>
#include <boost/multiprecision/random.hpp>
#include <iostream>

int main() 
{
    namespace mp = boost::multiprecision;
    using Int = mp::mpz_int;

    boost::mt19937 rng(3); // fixing seed for demo
    boost::uniform_int<Int> gen(-pow(Int(2), 400), pow(Int(2), 400));

    for (int i=0; i<10; ++i)
        std::cout << gen(rng) << "\n";
}

Prints:

-1933652715378833784248363501979652496795524829874743132697181322648247480952527478485970668716806865063045317090084841622
-1468881213423638668843172777487968633185322950671475476288214928834762957270366851213470028408822700452414112732095789399
-438410269130756764874038685663894375462191729266787898021001121529040612201593866121171654735148672212107859934777249455
1640218057904403949431629353470797958943403389857715009204662011172706206212175540336638682612917363014646204359229208161
2080556950904231221688486212902649050443577133350992840521950777901511719409216800649680002332270542856870490906586634021
-2462847552934789794366130512379986584363897268428607239076390917679673552257507232435012961043902569359791960286013555735
1862125165202309929540318106374963238582997665808535945941185531430178511983671609033768595314282085775703389782906055681
-2252919975572088150601736662143078753669379374770846936106371833826830834376177961242332270710557376868189820866644291936
986998873018694187216384502983789929097242088320473495018118860428802373488463609060400540575932015408503979156759366945
111638721010166959954664901006097000984357549061159193446548907668369849648549091048841517202745565156043524728780018634

---

^[1] For rationale see the header:

namespace boost{ namespace random{ namespace detail{
//
// This is a horrible hack: this declaration has to appear before the definition of
// uniform_int_distribution, otherwise it won't be used...
// Need to find a better solution, like make Boost.Random safe to use with
// UDT's and depricate/remove this header altogether.
//
template<class Engine, class Backend, boost::multiprecision::expression_template_option ExpressionTemplates>
boost::multiprecision::number<Backend, ExpressionTemplates> 
   generate_uniform_int(Engine& eng, const boost::multiprecision::number<Backend, ExpressionTemplates>& min_value, const boost::multiprecision::number<Backend, ExpressionTemplates>& max_value);

}}}

Answer 2

You can seed an independent_bits_engine in the following way:

gen.seed(a_value_used_as_seed)

See this .

EDIT

The mt19937 produces uint32_t integers and uses the same type as seed. However, it accepts also a seed sequence, wich can be used to create seeds that are beter distributed across the entire 32-bit range. The seed sequence can be any class, as long as it defines some functions. One of them is the following:

void generate(type*,type*)

These two different ways of seeding are represented by two template functions.

The type that is used in the independent_bits_engine is the mpz_int integer. When the engine is seeded, it passes the seed to the base engine, the mersenne twister. For the mt19937 the mpz_int , is not its base type, so the compiler selects the second template function and interprets the seed as a seed sequence. This causes an error because the mpz_int class, doesn't have a generator function.