PersistentThreadPool Class Reference

A generic thread pool used for intra-evaluation parallelism and batch processing. More...

#include <PersistentThreadPool.hpp>

Classes
struct	Job

Public Member Functions
	PersistentThreadPool (int workers)
	~PersistentThreadPool ()
	PersistentThreadPool (const PersistentThreadPool &)=delete
PersistentThreadPool &	operator= (const PersistentThreadPool &)=delete
void	parallel_for (int n, const std::function< void(int)> &fn)

Private Member Functions
void	worker_loop ()

Static Private Member Functions
static void	finish_one (const std::shared_ptr< Job > &job)
static void	run_job (const std::shared_ptr< Job > &job)

Private Attributes
std::vector< std::thread >	threads_
std::mutex	mu_
std::condition_variable	cv_
std::queue< std::shared_ptr< Job > >	jobs_
bool	stop_ = false

Detailed Description

A generic thread pool used for intra-evaluation parallelism and batch processing.

This thread pool is used by the MoEDoubleAccumulator in the core engine when the UCI EvalThreads option is set to >1. It parallelizes the evaluation of the neural network's internal spatial branches.

It is also used heavily by the standalone benchmark tools in src/tools/ to evaluate massive batches of positions simultaneously.

Constructor & Destructor Documentation

PersistentThreadPool() [1/2]

PersistentThreadPool::PersistentThreadPool ( int workers )

inlineexplicit

Here is the call graph for this function:

Here is the caller graph for this function:

~PersistentThreadPool()

PersistentThreadPool::~PersistentThreadPool ( )

inline

PersistentThreadPool() [2/2]

PersistentThreadPool::PersistentThreadPool ( const PersistentThreadPool & )

delete

Here is the call graph for this function:

Member Function Documentation

finish_one()

void PersistentThreadPool::finish_one ( const std::shared_ptr< Job > & job )

inlinestaticprivate

Here is the caller graph for this function:

operator=()

PersistentThreadPool & PersistentThreadPool::operator= ( const PersistentThreadPool & )

delete

Here is the call graph for this function:

parallel_for()

void PersistentThreadPool::parallel_for ( int n, const std::function< void(int)> & fn )

inline

Here is the call graph for this function:

run_job()

void PersistentThreadPool::run_job ( const std::shared_ptr< Job > & job )

inlinestaticprivate

Here is the call graph for this function:

Here is the caller graph for this function:

worker_loop()

void PersistentThreadPool::worker_loop ( )

inlineprivate

Here is the call graph for this function:

Here is the caller graph for this function:

Member Data Documentation

cv_

std::condition_variable PersistentThreadPool::cv_

private

jobs_

std::queue<std::shared_ptr<Job> > PersistentThreadPool::jobs_

private

mu_

std::mutex PersistentThreadPool::mu_

private

stop_

bool PersistentThreadPool::stop_ = false

private

threads_

std::vector<std::thread> PersistentThreadPool::threads_

private

---

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

• engine/src/eval/PersistentThreadPool.hpp