mirror of
https://github.com/triqs/dft_tools
synced 2024-11-01 19:53:45 +01:00
114 lines
4.9 KiB
C++
114 lines
4.9 KiB
C++
|
/*******************************************************************************
|
||
|
*
|
||
|
* TRIQS: a Toolbox for Research in Interacting Quantum Systems
|
||
|
*
|
||
|
* Copyright (C) 2013 by M. Ferrero, O. Parcollet, I. Krivenko
|
||
|
*
|
||
|
* TRIQS is free software: you can redistribute it and/or modify it under the
|
||
|
* terms of the GNU General Public License as published by the Free Software
|
||
|
* Foundation, either version 3 of the License, or (at your option) any later
|
||
|
* version.
|
||
|
*
|
||
|
* TRIQS is distributed in the hope that it will be useful, but WITHOUT ANY
|
||
|
* WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
|
||
|
* FOR A PARTICULAR PURPOSE. See the GNU General Public License for more
|
||
|
* details.
|
||
|
*
|
||
|
* You should have received a copy of the GNU General Public License along with
|
||
|
* TRIQS. If not, see <http://www.gnu.org/licenses/>.
|
||
|
*
|
||
|
******************************************************************************/
|
||
|
#ifndef TRIQS_CTQMC_KRYLOV_TIME_PT_HPP
|
||
|
#define TRIQS_CTQMC_KRYLOV_TIME_PT_HPP
|
||
|
|
||
|
#include <triqs/utility/first_include.hpp>
|
||
|
#include <limits>
|
||
|
#include <iostream>
|
||
|
#include <cmath>
|
||
|
|
||
|
namespace triqs { namespace utility {
|
||
|
|
||
|
/// A point on a very thin grid, as uint64_t
|
||
|
struct time_pt {
|
||
|
|
||
|
time_pt() { beta = 1; val =0; n = 0;}
|
||
|
explicit time_pt(double b) { beta = b; val =0; n = 0;}
|
||
|
|
||
|
private :
|
||
|
// too dangerous because of rounding to be left public
|
||
|
time_pt(double v, double beta_) { beta = beta_; val =v; n = std::floor(Nmax*(v/beta));}
|
||
|
time_pt(uint64_t n_, double beta_, bool) { beta = beta_; val = beta*(double(n_)/Nmax); n = n_;}
|
||
|
|
||
|
public :
|
||
|
|
||
|
// random case : to be improved, using rng only for integer for reproducibility....
|
||
|
template<typename RNG, typename T1, typename T2> static time_pt random(RNG & rng, T1 l, T2 beta_) { return time_pt(rng(double(l)), double(beta_));}
|
||
|
|
||
|
time_pt (time_pt const &) = default;
|
||
|
time_pt (time_pt && x) = default;
|
||
|
time_pt & operator = (time_pt const &) = default ;
|
||
|
time_pt & operator = (time_pt && x) = default;
|
||
|
|
||
|
//this is also dangerous for reproducibility
|
||
|
time_pt & operator = (double v) = delete;
|
||
|
|
||
|
// factories
|
||
|
static time_pt make_beta(double beta_) { time_pt r; r.beta = beta_; r.n = Nmax; r.val = beta_; return r;}
|
||
|
static time_pt make_zero(double beta_) { time_pt r; r.beta = beta_; r.n = 0; r.val = 0; return r;}
|
||
|
static time_pt make_from_double(double x, double beta_) { return time_pt(x,beta_);}
|
||
|
static time_pt epsilon(double beta) { return time_pt(1,beta,true);}
|
||
|
static time_pt epsilon(time_pt const & beta) { return time_pt(1,beta.beta,true);}
|
||
|
|
||
|
bool operator == (const time_pt & tp) const { return n == tp.n; }
|
||
|
bool operator != (const time_pt & tp) const { return n != tp.n; }
|
||
|
bool operator < (const time_pt & tp) const { return n < tp.n; }
|
||
|
bool operator <= (const time_pt & tp) const { return n <= tp.n; }
|
||
|
bool operator > (const time_pt & tp) const { return n > tp.n; }
|
||
|
bool operator >= (const time_pt & tp) const { return n >= tp.n; }
|
||
|
|
||
|
// adding and substracting is cyclic on [0, beta]
|
||
|
inline friend time_pt operator+(time_pt const & a, time_pt const & b) { return time_pt(a.n + b.n, a.beta, true); }
|
||
|
inline friend time_pt operator-(time_pt const & a, time_pt const & b) { uint64_t n = (a.n>= b.n ? a.n - b.n : Nmax - (b.n - a.n)); return time_pt(n, a.beta,true); }
|
||
|
|
||
|
//unary
|
||
|
inline friend time_pt operator-(time_pt const & a) { uint64_t n = Nmax - a.n; return time_pt(n, a.beta,true); }
|
||
|
|
||
|
// division by integer
|
||
|
inline friend time_pt div_by_int(time_pt const & a, size_t b) { return time_pt(a.n/ b, a.beta, true); }
|
||
|
|
||
|
// floor_div(x,y) = floor (x/y), but computed on the grid.
|
||
|
inline friend size_t floor_div(time_pt const & a, time_pt const & b){return a.n/b.n;}
|
||
|
|
||
|
// only EXPLICIT cast
|
||
|
explicit operator double() const {return val;} // cast to a double
|
||
|
|
||
|
friend std::ostream & operator<< (std::ostream & out, time_pt const & p) { return out << p.val << " [time_pt : beta = "<< p.beta<< " n = "<< p.n<<"]" ; }
|
||
|
|
||
|
private:
|
||
|
static constexpr uint64_t Nmax = std::numeric_limits<uint64_t>::max();
|
||
|
uint64_t n;
|
||
|
double val, beta;
|
||
|
};
|
||
|
|
||
|
// all operations below decay to double
|
||
|
inline double operator*(time_pt const & a, time_pt const & b) { return double(a)*double(b); }
|
||
|
inline double operator/(time_pt const & a, time_pt const & b) { return double(a)/double(b); }
|
||
|
|
||
|
#define IMPL_OP(OP) \
|
||
|
inline double operator OP(time_pt const & x, double y) {return static_cast<double>(x) OP y;} \
|
||
|
inline double operator OP(double y, time_pt const & x) {return y OP static_cast<double>(x);}
|
||
|
IMPL_OP(+); IMPL_OP(-); IMPL_OP(*); IMPL_OP(/);
|
||
|
#undef IMPL_OP
|
||
|
|
||
|
|
||
|
// all other operations : first cast into a double and do the operation
|
||
|
/*#define IMPL_OP(OP) \
|
||
|
template<typename T> auto operator OP(time_pt const & x, T y) -> decltype(double(0) OP y) {return static_cast<double>(x) OP y;} \
|
||
|
template<typename T> auto operator OP(T y, time_pt const & x) -> decltype(y OP double(0)) {return y OP static_cast<double>(x);} \
|
||
|
IMPL_OP(+); IMPL_OP(-); IMPL_OP(*); IMPL_OP(/);
|
||
|
#undef IMPL_OP*/
|
||
|
|
||
|
}}
|
||
|
#endif
|
||
|
|