duckdb::TaskScheduler Class Reference

The TaskScheduler is responsible for managing tasks and threads. More...

#include <duckdb.hpp>

Collaboration diagram for duckdb::TaskScheduler:

Public Member Functions
	TaskScheduler (DatabaseInstance &db)
unique_ptr< ProducerToken >	CreateProducer ()
void	ScheduleTask (ProducerToken &producer, shared_ptr< Task > task)
	Schedule a task to be executed by the task scheduler.
void	ScheduleTasks (ProducerToken &producer, vector< shared_ptr< Task > > &tasks)
bool	GetTaskFromProducer (ProducerToken &token, shared_ptr< Task > &task)
	Fetches a task from a specific producer, returns true if successful or false if no tasks were available.
void	ExecuteForever (atomic< bool > *marker)
	Run tasks forever until "marker" is set to false, "marker" must remain valid until the thread is joined.
idx_t	ExecuteTasks (atomic< bool > *marker, idx_t max_tasks)
void	ExecuteTasks (idx_t max_tasks)
	Run tasks until max_tasks have been completed, or until there are no more tasks available.
void	SetThreads (idx_t total_threads, idx_t external_threads)
void	RelaunchThreads ()
DUCKDB_API int32_t	NumberOfThreads ()
	Returns the number of threads.
idx_t	GetNumberOfTasks () const
idx_t	GetProducerCount () const
idx_t	GetTaskCountForProducer (ProducerToken &token) const
void	Signal (idx_t n)
	Send signals to n threads, signalling for them to wake up and attempt to execute a task.
void	SetAllocatorFlushTreshold (idx_t threshold)
	Set the allocator flush threshold.
void	SetAllocatorBackgroundThreads (bool enable)
	Sets the allocator background thread.

Static Public Member Functions
static DUCKDB_API TaskScheduler &	GetScheduler (ClientContext &context)
static DUCKDB_API TaskScheduler &	GetScheduler (DatabaseInstance &db)
static void	YieldThread ()
	Yield to other threads.
static idx_t	GetEstimatedCPUId ()

Private Member Functions
void	RelaunchThreadsInternal (int32_t n, bool destroy)

Private Attributes
DatabaseInstance &	db
unique_ptr< ConcurrentQueue >	queue
	The task queue.
mutex	thread_lock
	Lock for modifying the thread count.
vector< unique_ptr< SchedulerThread > >	threads
	The active background threads of the task scheduler.
vector< unique_ptr< atomic< bool > > >	markers
	Markers used by the various threads, if the markers are set to "false" the thread execution is stopped.
atomic< idx_t >	allocator_flush_threshold
	The threshold after which to flush the allocator after completing a task.
atomic< bool >	allocator_background_threads
	Whether allocator background threads are enabled.
atomic< int32_t >	requested_thread_count
	Requested thread count (set by the 'threads' setting)
atomic< int32_t >	current_thread_count
	The amount of threads currently running.

Static Private Attributes
static constexpr int64_t	TASK_TIMEOUT_USECS = 5000

Detailed Description

The TaskScheduler is responsible for managing tasks and threads.

Member Function Documentation

ExecuteTasks()

idx_t duckdb::TaskScheduler::ExecuteTasks	(	atomic< bool > *	marker,
		idx_t	max_tasks
	)

Run tasks until marker is set to false, max_tasks have been completed, or until there are no more tasks available. Returns the number of tasks that were completed.

SetThreads()

void duckdb::TaskScheduler::SetThreads	(	idx_t	total_threads,
		idx_t	external_threads
	)

Sets the amount of background threads to be used for execution, based on the number of total threads and the number of external threads. External threads, e.g. the main thread, will also be used for execution. Launches total_threads - external_threads background worker threads.

GetEstimatedCPUId()

static idx_t duckdb::TaskScheduler::GetEstimatedCPUId ( )

static

Get the number of the CPU on which the calling thread is currently executing. Fallback to calling thread id if CPU number is not available. Result do not need to be exact 'return 0' is a valid fallback strategy

---

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

• external/duckdb/duckdb.hpp