3
0
mirror of https://github.com/triqs/dft_tools synced 2024-11-01 03:33:50 +01:00
dft_tools/triqs/mc_tools/mc_generic.hpp
Olivier Parcollet 8c0bf592d0 Fix compilation of py_converter.
- py_converters in hpp files
- add guard to prevent inclusion in the wrapper (to avoid 2
  specialisation of the same py_converter).
2014-05-20 22:24:30 +02:00

231 lines
10 KiB
C++

/*******************************************************************************
*
* TRIQS: a Toolbox for Research in Interacting Quantum Systems
*
* Copyright (C) 2011-2013 by M. Ferrero, O. Parcollet
*
* 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/>.
*
******************************************************************************/
#pragma once
#include <triqs/utility/first_include.hpp>
#include <math.h>
#include <triqs/utility/timer.hpp>
#include <triqs/utility/report_stream.hpp>
#include <triqs/parameters/parameters.hpp>
#include "./mc_measure_aux_set.hpp"
#include "./mc_measure_set.hpp"
#include "./mc_move_set.hpp"
#include "./mc_basic_step.hpp"
#include "./random_generator.hpp"
namespace triqs { namespace mc_tools {
/**
* \brief Generic Monte Carlo class.
*/
template<typename MCSignType, typename MCStepType = Step::Metropolis<MCSignType> >
class mc_generic {
public:
/**
* Constructor from a set of parameters
*/
mc_generic(uint64_t N_Cycles, uint64_t Length_Cycle, uint64_t N_Warmup_Cycles, std::string Random_Name, int Random_Seed, int Verbosity,
std::function<bool()> AfterCycleDuty = std::function<bool()>() ) :
RandomGenerator(Random_Name, Random_Seed),
AllMoves(RandomGenerator),
AllMeasures(),AllMeasuresAux(),
report(&std::cout, Verbosity),
Length_MC_Cycle(Length_Cycle),
NWarmIterations(N_Warmup_Cycles),
NCycles(N_Cycles),
after_cycle_duty(AfterCycleDuty),
sign_av(0) {}
/**
* Constructor from a dictionnary
* \param[in] P dictionary parameters
* \param[in] AfterCycleDuty a function bool() to be called after each QMC cycle
*/
mc_generic(params::parameters const & P, std::function<bool()> AfterCycleDuty = std::function<bool()>() ) :
RandomGenerator(std::string(P["random_name"]), long(P["random_seed"])),
report(&std::cout,int(P["verbosity"])),
AllMoves(RandomGenerator),
AllMeasures(),AllMeasuresAux(),
Length_MC_Cycle(long(P["length_cycle"])), /// NOT NICE THIS EXPLICIT CAST : no unsigned in parameters, really ??
NWarmIterations(long(P["n_warmup_cycles"])),
NCycles(long(P["n_cycles"])),
after_cycle_duty(AfterCycleDuty),
sign_av(0) {}
/**
* Register move M with its probability of being proposed.
* NB : the proposition_probability needs to be >0 but does not need to be
* normalized. Normalization is automatically done with all the added moves
* before starting the run
*/
template <typename MoveType>
void add_move (MoveType && M, std::string name, double proposition_probability = 1.0) {
static_assert( !std::is_pointer<MoveType>::value, "add_move in mc_generic takes ONLY values !");
AllMoves.add(std::forward<MoveType>(M), name, proposition_probability);
}
/**
* Register the Measure M
*/
template<typename MeasureType>
void add_measure (MeasureType && M, std::string name) {
static_assert( !std::is_pointer<MeasureType>::value, "add_measure in mc_generic takes ONLY values !");
AllMeasures.insert(std::forward<MeasureType>(M), name);
}
/**
* Register the precomputation
*/
template <typename MeasureAuxType> void add_measure_aux(std::shared_ptr<MeasureAuxType> p) { AllMeasuresAux.emplace_back(p); }
/// get the average sign (to be called after collect_results)
MCSignType average_sign() const { return sign_av; }
/// get the current percents done
uint64_t percent() const { return done_percent; }
// An access to the random number generator
random_generator & rng(){ return RandomGenerator;}
/// Start the Monte Carlo
bool start(MCSignType sign_init, std::function<bool ()> const & stop_callback) {
assert(stop_callback);
Timer.start();
sign = sign_init; done_percent = 0; nmeasures = 0;
sum_sign = 0;
bool stop_it=false, finished = false;
uint64_t NCycles_tot = NCycles+ NWarmIterations;
report << std::endl << std::flush;
for (NC =0; !stop_it; ++NC) {
for (uint64_t k=1; (k<=Length_MC_Cycle); k++) { MCStepType::do_it(AllMoves,RandomGenerator,sign); }
if (after_cycle_duty) {after_cycle_duty();}
if (thermalized()) {
nmeasures++;
sum_sign += sign;
for (auto &x : AllMeasuresAux) x();
AllMeasures.accumulate(sign);
}
// recompute fraction done
uint64_t dp = uint64_t(floor( ( NC*100.0) / (NCycles_tot-1)));
if (dp>done_percent) { done_percent=dp; report << done_percent; report<<"%; "; report <<std::flush; }
finished = ( (NC >= NCycles_tot -1) || converged () );
stop_it = (stop_callback() || finished);
}
report << std::endl << std::endl << std::flush;
Timer.stop();
return finished;
}
/// Reduce the results of the measures, and reports some statistics
void collect_results(boost::mpi::communicator const & c) {
uint64_t nmeasures_tot;
MCSignType sum_sign_tot;
boost::mpi::reduce(c, nmeasures, nmeasures_tot, std::plus<uint64_t>(), 0);
boost::mpi::reduce(c, sum_sign, sum_sign_tot, std::plus<MCSignType>(), 0);
report(3) << "[Node "<<c.rank()<<"] Acceptance rate for all moves:\n" << AllMoves.get_statistics(c);
report(3) << "[Node "<<c.rank()<<"] Simulation lasted: " << double(Timer) << " seconds" << std::endl;
report(3) << "[Node "<<c.rank()<<"] Number of measures: " << nmeasures << std::endl;
report(3) << "[Node "<<c.rank()<<"] Average sign: " << sum_sign / double(nmeasures) << std::endl << std::endl << std::flush;
if (c.rank()==0) {
sign_av = sum_sign_tot / double(nmeasures_tot);
report(2) << "Total number of measures: " << nmeasures_tot << std::endl;
report(2) << "Average sign: " << sign_av << std::endl << std::endl << std::flush;
}
AllMeasures.collect_results(c);
}
// do not use direcly, use the free function it is simpler to call...
template<typename MeasureType> MeasureType & get_measure(std::string const & name) { return AllMeasures.template get_measure<MeasureType> (name); }
template<typename MeasureType> MeasureType const & get_measure(std::string const & name) const { return AllMeasures.template get_measure<MeasureType> (name); }
template<typename MoveType> MoveType & get_move (std::string const & name) { return AllMoves.template get_move<MoveType> (name); }
template<typename MoveType> MoveType const & get_move (std::string const & name) const { return AllMoves.template get_move<MoveType> (name); }
/// HDF5 interface
friend void h5_write (h5::group g, std::string const & name, mc_generic const & mc){
auto gr = g.create_group(name);
h5_write(gr,"moves", mc.AllMoves);
h5_write(gr,"measures", mc.AllMeasures);
h5_write(gr,"length_monte_carlo_cycle", mc.Length_MC_Cycle);
h5_write(gr,"number_cycle_requested", mc.NCycles);
h5_write(gr,"number_warming_cycle_requested", mc.NWarmIterations);
h5_write(gr,"number_cycle_done", mc.NC);
h5_write(gr,"number_measure_done", mc.nmeasures);
h5_write(gr,"sign", mc.sign);
h5_write(gr,"sum_sign", mc.sum_sign);
}
/// HDF5 interface
friend void h5_read (h5::group g, std::string const & name, mc_generic & mc){
auto gr = g.open_group(name);
h5_read(gr,"moves", mc.AllMoves);
h5_read(gr,"measures", mc.AllMeasures);
h5_read(gr,"length_monte_carlo_cycle", mc.Length_MC_Cycle);
h5_read(gr,"number_cycle_requested", mc.NCycles);
h5_read(gr,"number_warming_cycle_requested", mc.NWarmIterations);
h5_read(gr,"number_cycle_done", mc.NC);
h5_read(gr,"number_measure_done", mc.nmeasures);
h5_read(gr,"sign", mc.sign);
h5_read(gr,"sum_sign", mc.sum_sign);
}
private:
random_generator RandomGenerator;
move_set<MCSignType> AllMoves;
measure_set<MCSignType> AllMeasures;
std::vector<measure_aux> AllMeasuresAux;
utility::report_stream report;
uint64_t Length_MC_Cycle;/// Length of one Monte-Carlo cycle between 2 measures
uint64_t NWarmIterations, NCycles;
uint64_t nmeasures;
MCSignType sum_sign;
utility::timer Timer;
std::function<bool()> after_cycle_duty;
MCSignType sign, sign_av;
uint64_t NC,done_percent;// NC = number of the cycle
bool thermalized() const { return (NC>= NWarmIterations);}
bool converged() const { return false;}
};
/// Retrieve a Measure given name and type. NB : the type is checked at runtime
template<typename M,typename T1, typename T2> M & get_measure(mc_generic<T1,T2> & s, std::string const & name) { return s.template get_measure<M> (name); }
template<typename M,typename T1, typename T2> M const & get_measure(mc_generic<T1,T2> const & s, std::string const & name) { return s.template get_measure<M> (name); }
/// Retrieve a Measure given name and type. NB : the type is checked at runtime
template<typename M,typename T1, typename T2> M * get_measure_aux(mc_generic<T1,T2> & s, std::string const & name) { return s.template get_measure_aux<M> (name); }
template<typename M,typename T1, typename T2> M const * get_measure_aux(mc_generic<T1,T2> const & s, std::string const & name) { return s.template get_measure_aux<M> (name); }
/// Retrieve a Move given name and type. NB : the type is checked at runtime
template<typename M,typename T1, typename T2> M & get_move(mc_generic<T1,T2> & s, std::string const & name) { return s.template get_move<M> (name); }
template<typename M,typename T1, typename T2> M const & get_move(mc_generic<T1,T2> const & s, std::string const & name) { return s.template get_move<M> (name); }
}}// end namespace