cps::AlgHmcPhi Class Reference
[Algorithms]

A class implementing the Hybrid Monte Carlo algorithm. More...

#include <alg_hmd.h>

Inherits cps::AlgHmd.

List of all members.

Public Member Functions
	AlgHmcPhi (Lattice &latt, CommonArg *c_arg, HmdArg *arg)
virtual	~AlgHmcPhi ()
Float	run (void)
	Performs a single HMC trajectory.
Protected Attributes
int	n_frm_masses
	The number of dynamical fermion masses.
int	n_bsn_masses
	The number of dynamical boson masses.
int	f_size
	The size of the pseudofermion (and similar) fields.
CgArg **	frm_cg_arg
	Pointer to an array of structures containing solver parameters.
CgArg **	bsn_cg_arg
	Pointer to an array of structures containing solver parameters.
Vector **	phi
	Pseudofermion fields.
Vector **	bsn
	Boson (pseudoboson?) fields.
Matrix *	gauge_field_init
	The initial gauge field configuration.
Vector **	frm1
	Array of general purpose fields.
Vector **	frm2
	Another array of general purpose fields.
Vector **	cg_sol_cur
	Pointer to the most recent solution produced by the solver.
Vector **	cg_sol_prev
	Pointer to the most recent solution produced by the solver.
Float *	h_f_init
	The initial value of the pseudofermion action.
Float *	h_f_final
	The final value of the pseudofermion action.
Float *	delta_h_f
	The change in the value of the pseudofermion action.
Float *	h_b_init
	The initial value of the boson action.
Float *	h_b_final
	The final value of the boson action.
Float *	delta_h_b
	The change in the value of the boson action.

---

Detailed Description

A class implementing the Hybrid Monte Carlo algorithm.

This evolves a gauge field by a single iteration of the standard HMC algorithm (phi algorithm), i.e. a molecular dynamics trajectory followed by a metropolis accept/reject step.

This implementation uses a Upqp leapfrog integration scheme and a two-step chronological method to choose the starting guess for the solver.

The algorithm is configurable to include dynamical fermions of several masses, each mass having its own set of pseudofermion fields and solver parameters. The number of degenerate flavours at each mass depends on the type of fermion action used. Similarly one can have in addition pseudobosonic terms in the hamiltonian contributing to the total force on the momentum field.

One also has the option of forcing acceptance at the metropolis step.

A bunch of statistics relating to the performance of the solver, the conservation of the hamiltonian and the metropolis acceptance are collected.

Definition at line 99 of file alg_hmd.h.

---

Constructor & Destructor Documentation

cps::AlgHmcPhi::AlgHmcPhi (  Lattice &  latt, CommonArg *  c_arg, HmdArg *  arg )

Parameters: latt  The lattice on which the HMC algorithm runs. c_arg  The common argument structure for all algorithms. arg  The algorithm parameters. Definition at line 61 of file alg_hmc_phi.C. References bsn, bsn_cg_arg, cps::HmdArg::bsn_mass, cg_sol_cur, cg_sol_prev, cps::ERR, f_size, cps::Lattice::FchkbEvl(), cps::Float, frm1, frm2, frm_cg_arg, cps::HmdArg::frm_mass, cps::Lattice::FsiteSize(), cps::Verbose::Func(), gauge_field_init, cps::Error::General(), cps::GJP, cps::CgArg::mass, cps::MAX_HMD_MASSES, cps::HmdArg::max_num_iter, cps::CgArg::max_num_iter, cps::HmdArg::n_bsn_masses, n_bsn_masses, cps::HmdArg::n_frm_masses, n_frm_masses, phi, cps::Error::Pointer(), cps::Verbose::Smalloc(), cps::smalloc(), cps::HmdArg::stop_rsd, cps::CgArg::stop_rsd, cps::GlobalJobParameter::VolNodeSites(), and cps::VRB.

cps::AlgHmcPhi::~AlgHmcPhi (   )  [virtual]

Definition at line 248 of file alg_hmc_phi.C. References bsn, bsn_cg_arg, cg_sol_cur, cg_sol_prev, frm1, frm2, frm_cg_arg, cps::Verbose::Func(), cps::Verbose::FuncEnd(), gauge_field_init, n_bsn_masses, n_frm_masses, phi, cps::Verbose::Sfree(), cps::sfree(), and cps::VRB.

---

Member Function Documentation

Float cps::AlgHmcPhi::run (  void   )  [virtual]

Performs a single HMC trajectory. Returns: The probability used in the metropolis step. Postcondition: The following results are written to the file specified in the CommonArg structure: Number of molecular dynamics steps + 1 Change in the hamiltonian 1/0 if accepted/rejected A measure of the change in the gauge field due to reunitarisation Another measure of the change in the gauge field due to reunitarisation Average number of solver iterations. Minimum number of solver iterations. Maximum number of solver iterations. Average solver residue Minimum solver residue Maximum solver residue Implements cps::AlgHmd. Definition at line 362 of file alg_hmc_phi.C. References cps::Alg::AlgLattice(), cps::Lattice::BhamiltonNode(), bsn, bsn_cg_arg, cps::HmdArg::bsn_mass, cg_sol_cur, cg_sol_prev, cps::CNV_FRM_NO, cps::Lattice::CopyGaugeField(), cps::Vector::CopyVec(), cps::CSM, cps::ERR, cps::Lattice::EvolveGfield(), cps::Lattice::EvolveMomFforce(), cps::Lattice::EvolveMomGforce(), f_size, cps::Fclose(), cps::Lattice::FhamiltonNode(), cps::Error::FileA(), cps::Float, cps::Verbose::Flow(), cps::Lattice::FmatEvlInv(), cps::Fopen(), cps::Fprintf(), frm1, frm2, frm_cg_arg, cps::HmdArg::frm_mass, cps::Vector::FTimesV1MinusV2(), cps::Verbose::Func(), gauge_field_init, cps::Lattice::GaugeField(), cps::Lattice::GhamiltonNode(), cps::GJP, cps::glb_sum(), cps::Lattice::GsoCheck(), cps::Lattice::GupdCnt(), cps::Lattice::GupdCntInc(), cps::Error::Hardware(), cps::IFloat, cps::Lattice::MdTime(), cps::Lattice::MdTimeInc(), cps::HmdArg::metropolis, cps::Lattice::MetropolisAccept(), cps::Lattice::MomHamiltonNode(), phi, cps::MPISCU::printf(), cps::Lattice::RandGaussAntiHermMatrix(), cps::Lattice::RandGaussVector(), cps::Verbose::Result(), cps::CommonArg::results, cps::Lattice::Reunitarize(), cps::HmdArg::reunitarize, cps::CheckSum::SaveComment(), cps::Lattice::SetPhi(), cps::GlobalJobParameter::Snodes(), cps::Lattice::SoCheck(), cps::HmdArg::step_size, cps::HmdArg::steps_per_traj, and cps::VRB.

---

Member Data Documentation

Vector** cps::AlgHmcPhi::bsn [protected]

Boson (pseudoboson?) fields. One for each mass Definition at line 142 of file alg_hmd.h. Referenced by AlgHmcPhi(), run(), and ~AlgHmcPhi().

CgArg** cps::AlgHmcPhi::bsn_cg_arg [protected]

Pointer to an array of structures containing solver parameters. These are the parameters corresponding to each of the dynamical boson masses. Definition at line 131 of file alg_hmd.h. Referenced by AlgHmcPhi(), run(), and ~AlgHmcPhi().

Vector** cps::AlgHmcPhi::cg_sol_cur [protected]

Pointer to the most recent solution produced by the solver. One for each mass Definition at line 154 of file alg_hmd.h. Referenced by AlgHmcPhi(), run(), and ~AlgHmcPhi().

Vector** cps::AlgHmcPhi::cg_sol_prev [protected]

Pointer to the most recent solution produced by the solver. One for each mass Definition at line 157 of file alg_hmd.h. Referenced by AlgHmcPhi(), run(), and ~AlgHmcPhi().

Float* cps::AlgHmcPhi::delta_h_b [protected]

The change in the value of the boson action. The final value - the initial value. One for each mass Definition at line 182 of file alg_hmd.h.

Float* cps::AlgHmcPhi::delta_h_f [protected]

The change in the value of the pseudofermion action. The final value - the initial value. One for each mass Definition at line 170 of file alg_hmd.h.

int cps::AlgHmcPhi::f_size [protected]

The size of the pseudofermion (and similar) fields. The size is given in terms of the total number of floating point numbers in the field on the local lattice, taking into account whether or not the field is defined on just a single parity. Definition at line 117 of file alg_hmd.h. Referenced by AlgHmcPhi(), and run().

Vector** cps::AlgHmcPhi::frm1 [protected]

Array of general purpose fields. Definition at line 149 of file alg_hmd.h. Referenced by AlgHmcPhi(), run(), and ~AlgHmcPhi().

Vector** cps::AlgHmcPhi::frm2 [protected]

Another array of general purpose fields. Definition at line 151 of file alg_hmd.h. Referenced by AlgHmcPhi(), run(), and ~AlgHmcPhi().

CgArg** cps::AlgHmcPhi::frm_cg_arg [protected]

Pointer to an array of structures containing solver parameters. These are the parameters corresponding to each of the dynamical fermion masses. Definition at line 124 of file alg_hmd.h. Referenced by AlgHmcPhi(), run(), and ~AlgHmcPhi().

Matrix* cps::AlgHmcPhi::gauge_field_init [protected]

The initial gauge field configuration. Definition at line 146 of file alg_hmd.h. Referenced by AlgHmcPhi(), run(), and ~AlgHmcPhi().

Float* cps::AlgHmcPhi::h_b_final [protected]

The final value of the boson action. The value at the end of the trajectory. One for each mass Definition at line 178 of file alg_hmd.h.

Float* cps::AlgHmcPhi::h_b_init [protected]

The initial value of the boson action. The value at the start of the trajectory. One for each mass Definition at line 174 of file alg_hmd.h.

Float* cps::AlgHmcPhi::h_f_final [protected]

The final value of the pseudofermion action. The value at the end of the trajectory. One for each mass Definition at line 166 of file alg_hmd.h.

Float* cps::AlgHmcPhi::h_f_init [protected]

The initial value of the pseudofermion action. The value at the start of the trajectory. One for each mass Definition at line 162 of file alg_hmd.h.

int cps::AlgHmcPhi::n_bsn_masses [protected]

The number of dynamical boson masses. This is not necessarily the number of dynamical flavours. Definition at line 114 of file alg_hmd.h. Referenced by AlgHmcPhi(), and ~AlgHmcPhi().

int cps::AlgHmcPhi::n_frm_masses [protected]

The number of dynamical fermion masses. Definition at line 108 of file alg_hmd.h. Referenced by AlgHmcPhi(), and ~AlgHmcPhi().

Vector** cps::AlgHmcPhi::phi [protected]

Pseudofermion fields. One for each mass Definition at line 138 of file alg_hmd.h. Referenced by AlgHmcPhi(), run(), and ~AlgHmcPhi().

---

The documentation for this class was generated from the following files:

• include/alg/alg_hmd.h
• src/alg/alg_hmd/alg_hmc_phi.C

---