ChainPool< HYP, Chain_t > Class Template Reference

#include <ChainPool.h>

Inheritance diagram for ChainPool< HYP, Chain_t >:

Collaboration diagram for ChainPool< HYP, Chain_t >:

Public Types
enum	RunningState { RunningState::READY, RunningState::RUNNING, RunningState::DONE }

Public Member Functions
	ChainPool ()
	ChainPool (HYP &h0, typename HYP::data_t *d, size_t n)
void	set_data (typename HYP::data_t *d, bool recompute=true)
	Set this data. More...
template<typename... ARGS>
void	add_chain (ARGS... args)
	Lock and modify the pool. More...
size_t	nchains () const
void	show (std::string prefix) const
generator< HYP & >	run_thread (Control &ctl) override
	This run helper is called internally by multiple different threads, and runs a given pool. More...
Public Member Functions inherited from ThreadedInferenceInterface< HYP >
virtual generator< HYP &>	run_thread (Control &ctl, Args... args)=0
	ThreadedInferenceInterface ()
unsigned long	next_index ()
	Return the next index to operate on (in a thread-safe way). More...
size_t	nthreads ()
	How many threads are currently run in this interface? More...
void	run_thread_generator_wrapper (size_t thr, Control &ctl, Args... args)
	We have to wrap run_thread in something that manages the sync with main. This really just synchronizes the output of run_thread with run below. NOTE this makes a copy of x into the local next_x, so that when the thread keeps running, it doesn't mess anything up. We may in the future block the thread and return a reference, but its not clear that's faster. More...
generator< HYP &>	run (Control ctl, Args... args)
	Set up the multiple threads and actually run, calling run_thread_generator_wrapper. More...
generator< HYP &>	unthreaded_run (Control ctl, Args... args)

Public Attributes
std::vector< Chain_t >	pool
unsigned long	steps_before_change = 100
std::vector< RunningState >	running
OrderedLock	running_lock
Public Attributes inherited from ThreadedInferenceInterface< HYP >
std::atomic< size_t >	index
size_t	__nthreads
std::atomic< size_t >	__nrunning
ConcurrentQueue< HYP >	to_yield

Detailed Description

template<typename HYP, typename Chain_t = MCMCChain<HYP>>
class ChainPool< HYP, Chain_t >

Author

steven piantadosi

Date

29/01/20

Member Enumeration Documentation

RunningState

template<typename HYP, typename Chain_t = MCMCChain<HYP>>

enum ChainPool::RunningState

strong

Enumerator
READY
RUNNING
DONE

Constructor & Destructor Documentation

ChainPool() [1/2]

template<typename HYP, typename Chain_t = MCMCChain<HYP>>

ChainPool< HYP, Chain_t >::ChainPool ( )

inline

ChainPool() [2/2]

template<typename HYP, typename Chain_t = MCMCChain<HYP>>

ChainPool< HYP, Chain_t >::ChainPool ( HYP &  h0, typename HYP::data_t *  d, size_t  n  )

inline

Member Function Documentation

add_chain()

template<typename HYP, typename Chain_t = MCMCChain<HYP>>

template<typename... ARGS>

void ChainPool< HYP, Chain_t >::add_chain ( ARGS...  args )

inline

Lock and modify the pool.

Parameters

h0
d

nchains()

template<typename HYP, typename Chain_t = MCMCChain<HYP>>

size_t ChainPool< HYP, Chain_t >::nchains ( ) const

inline

run_thread()

template<typename HYP, typename Chain_t = MCMCChain<HYP>>

generator<HYP&> ChainPool< HYP, Chain_t >::run_thread ( Control &  ctl )

inlineoverride

This run helper is called internally by multiple different threads, and runs a given pool.

Parameters

ctl

figure out how many steps to run

set_data()

template<typename HYP, typename Chain_t = MCMCChain<HYP>>

void ChainPool< HYP, Chain_t >::set_data ( typename HYP::data_t *  d, bool  recompute = true  )

inline

Set this data.

Parameters

d	– what data to set
recompute	– should I recompute all of the posteriors?

show()

template<typename HYP, typename Chain_t = MCMCChain<HYP>>

void ChainPool< HYP, Chain_t >::show ( std::string  prefix ) const

inline

Member Data Documentation

pool

template<typename HYP, typename Chain_t = MCMCChain<HYP>>

std::vector<Chain_t> ChainPool< HYP, Chain_t >::pool

running

template<typename HYP, typename Chain_t = MCMCChain<HYP>>

std::vector<RunningState> ChainPool< HYP, Chain_t >::running

running_lock

template<typename HYP, typename Chain_t = MCMCChain<HYP>>

OrderedLock ChainPool< HYP, Chain_t >::running_lock

steps_before_change

template<typename HYP, typename Chain_t = MCMCChain<HYP>>

unsigned long ChainPool< HYP, Chain_t >::steps_before_change = 100

---

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

• src/Inference/MCMC/ChainPool.h