Statistics Class Reference

Running time averages per owned cell: means of fields and covariances <a'b'> of pairs of them, updated on the device every interval steps once t > t_start. A sample at time t has weight t - t_prev (the previous sample's time, or t_start), and with a = w / W over the total weight W the update is (West 1979) d_a = a_new - mean_a, mean_a += a d_a, cov_ab += a (d_a (b_new - mean_b,new) - cov_ab), so each output field is the current average itself. Restart files carry the averages and the weights, so a restarted run continues them bitwise. More...

#include <statistics.h>

Public Member Functions
void	init (const toml::value &input, const std::vector< std::string > &species_names, uint32_t n_cells)
	Read [statistics] (if any) and allocate n_cells.
bool	enabled () const
bool	due (uint64_t step, rtype t) const
void	sample (rtype t, const CellSampler &sampler, uint32_t n_owned)
	Add the sample of time t of cells [0, n_owned).
std::vector< std::string >	variables () const
	Output variables: MEAN_<field>..., COV__...
void	register_data (std::vector< Data > &data) const
void	copy_device_to_host ()
void	restore (const RestartData &restart, const std::string &file, rtype t)
	Continue the averages of a restart file, or start afresh at time t if it holds none (or with reset); throws if it holds only some of them.
void	start (rtype t)
	Start afresh at time t (first sample weighted from max(t, t_start)).
RestartAttributes	attributes () const
std::string	summary () const
uint64_t	samples () const
const auto &	host_means () const
	Host copies (see copy_device_to_host): (cell, field) means, (cell, product) covariances.
const auto &	host_covariances () const

Detailed Description

Running time averages per owned cell: means of fields and covariances <a'b'> of pairs of them, updated on the device every interval steps once t > t_start. A sample at time t has weight t - t_prev (the previous sample's time, or t_start), and with a = w / W over the total weight W the update is (West 1979) d_a = a_new - mean_a, mean_a += a d_a, cov_ab += a (d_a (b_new - mean_b,new) - cov_ab), so each output field is the current average itself. Restart files carry the averages and the weights, so a restarted run continues them bitwise.

Member Function Documentation

attributes()

RestartAttributes Statistics::attributes

(

)

const

copy_device_to_host()

void Statistics::copy_device_to_host

(

)

due()

bool Statistics::due ( uint64_t  step, rtype  t  ) const

inline

enabled()

bool Statistics::enabled ( ) const

inline

host_covariances()

const auto & Statistics::host_covariances ( ) const

inline

host_means()

const auto & Statistics::host_means ( ) const

inline

Host copies (see copy_device_to_host): (cell, field) means, (cell, product) covariances.

init()

void Statistics::init	(	const toml::value &	input,
		const std::vector< std::string > &	species_names,
		uint32_t	n_cells
	)

Read [statistics] (if any) and allocate n_cells.

register_data()

void Statistics::register_data

(

std::vector< Data > &

data

)

const

restore()

void Statistics::restore	(	const RestartData &	restart,
		const std::string &	file,
		rtype	t
	)

Continue the averages of a restart file, or start afresh at time t if it holds none (or with reset); throws if it holds only some of them.

sample()

void Statistics::sample	(	rtype	t,
		const CellSampler &	sampler,
		uint32_t	n_owned
	)

Add the sample of time t of cells [0, n_owned).

samples()

uint64_t Statistics::samples ( ) const

inline

start()

void Statistics::start

(

rtype

t

)

Start afresh at time t (first sample weighted from max(t, t_start)).

summary()

std::string Statistics::summary

(

)

const

variables()

std::vector< std::string > Statistics::variables

(

)

const

Output variables: MEAN_<field>..., COV__...

---

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

• /home/runner/work/mallard/mallard/mallard/src/solver/statistics.h
• /home/runner/work/mallard/mallard/mallard/src/solver/statistics.cpp