1
0
mirror of https://github.com/VCMP-SqMod/SqMod.git synced 2024-11-08 08:47:17 +01:00
SqMod/include/RandomLib/MPFRExponentialL.hpp

124 lines
4.1 KiB
C++

/**
* \file MPFRExponentialL.hpp
* \brief Header for MPFRExponentialL
*
* Sampling exactly from the exponential distribution for MPFR using the
* traditional method.
*
* Copyright (c) Charles Karney (2012) <charles@karney.com> and licensed under
* the MIT/X11 License. For more information, see
* http://randomlib.sourceforge.net/
**********************************************************************/
#if !defined(RANDOMLIB_MPFREXPONENTIALL_HPP)
#define RANDOMLIB_MPFREXPONENTIALL_HPP 1
#include <cmath> // for log
#include <mpfr.h>
#define HAVE_MPFR (MPFR_VERSION_MAJOR >= 3)
#if HAVE_MPFR || defined(DOXYGEN)
namespace RandomLib {
/**
* \brief The exponential distribution for MPFR (the log method).
*
* This class is <b>DEPRECATED</b>. It is included for illustrative purposes
* only. The MPFRExponential class provides a much more efficient method for
* sampling from the exponential distribution.
*
* This is an adaption of ExponentialDistribution to MPFR. The changes are
* - Use MPFR's random number generator
* - Use sufficient precision internally to ensure that a correctly rounded
* result is returned.
*
* This class uses mutable private objects. So a single MPFRExponentialL
* object cannot safely be used by multiple threads. In a multi-processing
* environment, each thread should use a thread-specific MPFRExponentialL
* object.
**********************************************************************/
class MPFRExponentialL {
private:
// The number of bits of randomness to add at a time.
static const long chunk_ = 32;
public:
/**
* Initialize the MPFRExponentialL object.
**********************************************************************/
MPFRExponentialL() {
mpz_init(_vi);
mpfr_init2(_eps, chunk_);
mpfr_init2(_v1, chunk_);
mpfr_init2(_v2, chunk_);
}
/**
* Destroy the MPFRExponentialL object.
**********************************************************************/
~MPFRExponentialL() {
mpfr_clear(_v2);
mpfr_clear(_v1);
mpfr_clear(_eps);
mpz_clear(_vi);
}
/**
* Sample from the exponential distribution with mean 1.
*
* @param[out] val the sample from the exponential distribution
* @param[in,out] r a GMP random generator.
* @param[in] round the rounding direction.
* @return the MPFR ternary result (&plusmn; if val is larger/smaller than
* the exact sample).
**********************************************************************/
int operator()(mpfr_t val, gmp_randstate_t r, mpfr_rnd_t round) const {
mpfr_prec_t prec0 = mpfr_get_prec (val);
mpfr_prec_t prec = prec0 + 10; // A rough optimum
mpz_urandomb(_vi, r, prec);
mpfr_set_ui_2exp(_eps, 1u, -prec, MPFR_RNDN);
mpfr_set_prec(_v1, prec);
mpfr_set_z_2exp(_v1, _vi, -prec, MPFR_RNDN);
mpfr_set_prec(_v2, prec);
mpfr_add(_v2, _v1, _eps, MPFR_RNDN);
while (true) {
int f2 = mpfr_log(val, _v2, round); // val = log(upper bound)
mpfr_set_prec(_v2, prec0);
int f1 = mpfr_log(_v2, _v1, round); // v2 = log(lower bound)
if (f1 == f2 && mpfr_equal_p(val, _v2)) {
mpfr_neg(val, val, MPFR_RNDN);
return -f1;
}
prec = Refine(r, prec);
}
}
private:
// disable copy constructor and assignment operator
MPFRExponentialL(const MPFRExponentialL&);
MPFRExponentialL& operator=(const MPFRExponentialL&);
// Refine the random interval
mpfr_prec_t Refine(gmp_randstate_t r, mpfr_prec_t prec)
const {
prec += chunk_;
mpfr_div_2ui(_eps, _eps, chunk_, MPFR_RNDN);
mpz_urandomb(_vi, r, chunk_);
mpfr_set_prec(_v2, prec);
mpfr_set_z_2exp(_v2, _vi, -prec, MPFR_RNDN);
mpfr_add(_v2, _v1, _v2, MPFR_RNDN);
mpfr_swap(_v1, _v2); // v1 = v2;
mpfr_set_prec(_v2, prec);
mpfr_add(_v2, _v1, _eps, MPFR_RNDN);
return prec;
}
mutable mpz_t _vi;
mutable mpfr_t _eps;
mutable mpfr_t _v1;
mutable mpfr_t _v2;
};
} // namespace RandomLib
#endif // HAVE_MPFR
#endif // RANDOMLIB_MPFREXPONENTIALL_HPP