the workscheduler More...

#include <COM_WorkScheduler.h>

Static Public Member Functions
Public API
static void	schedule (WorkPackage *package)
	schedule a chunk of a group to be calculated. An execution group schedules a chunk in the WorkScheduler when ExecutionGroup.get_flags().open_cl is set the work will be handled by a OpenCLDevice otherwise the work is scheduled for an CPUDevice More...

static void	initialize (bool use_opencl, int num_cpu_threads)
	initialize the WorkScheduler More...

static void	deinitialize ()
	deinitialize the WorkScheduler free all allocated resources More...

static void	start (CompositorContext &context)
	Start the execution this methods will start the WorkScheduler. Inside this method all threads are initialized. for every device a thread is created. More...

static void	stop ()
	stop the execution All created thread by the start method are destroyed. More...

static void	finish ()
	wait for all work to be completed. More...

static bool	has_gpu_devices ()
	Are there OpenCL capable GPU devices initialized? the result of this method is stored in the CompositorContext A node can generate a different operation tree when OpenCLDevices exists. More...

static int	current_thread_id ()

Detailed Description

the workscheduler

Definition at line 32 of file COM_WorkScheduler.h.

Member Function Documentation

◆ current_thread_id()

int blender::compositor::WorkScheduler::current_thread_id ( )

static

Definition at line 571 of file COM_WorkScheduler.cc.

References BLI_thread_local_get, blender::compositor::COM_threading_model(), blender::compositor::SingleThreaded, and blender::compositor::CPUDevice::thread_id().

Referenced by blender::compositor::TextureBaseOperation::executePixelSampled().

◆ deinitialize()

void blender::compositor::WorkScheduler::deinitialize ( )

static

deinitialize the WorkScheduler free all allocated resources

Definition at line 550 of file COM_WorkScheduler.cc.

References blender::compositor::COM_is_opencl_enabled(), blender::compositor::COM_threading_model(), blender::compositor::opencl_deinitialize(), blender::compositor::Queue, blender::compositor::SingleThreaded, blender::compositor::Task, and blender::compositor::threading_model_queue_deinitialize().

Referenced by COM_deinitialize().

◆ finish()

void blender::compositor::WorkScheduler::finish ( )

static

wait for all work to be completed.

Definition at line 479 of file COM_WorkScheduler.cc.

References blender::compositor::COM_is_opencl_enabled(), blender::compositor::COM_threading_model(), blender::compositor::opencl_finish(), blender::compositor::Queue, blender::compositor::SingleThreaded, blender::compositor::Task, blender::compositor::threading_model_queue_finish(), and blender::compositor::threading_model_task_finish().

Referenced by blender::compositor::ExecutionSystem::execute(), and blender::compositor::ExecutionGroup::execute().

◆ has_gpu_devices()

bool blender::compositor::WorkScheduler::has_gpu_devices ( )

static

Are there OpenCL capable GPU devices initialized? the result of this method is stored in the CompositorContext A node can generate a different operation tree when OpenCLDevices exists.

See also: CompositorContext.getHasActiveOpenCLDevices

Definition at line 521 of file COM_WorkScheduler.cc.

References blender::compositor::COM_is_opencl_enabled(), and blender::compositor::opencl_has_gpu_devices().

Referenced by blender::compositor::ExecutionSystem::ExecutionSystem().

◆ initialize()

void blender::compositor::WorkScheduler::initialize	(	bool	use_opencl,
		int	num_cpu_threads
	)

static

initialize the WorkScheduler

during initialization the mutexes are initialized. there are two mutexes (for every device type one) After mutex initialization the system is queried in order to count the number of CPUDevices and GPUDevices to be created. For every hardware thread a CPUDevice and for every OpenCL GPU device a OpenCLDevice is created. these devices are stored in a separate list (cpudevices & gpudevices)

This function can be called multiple times to lazily initialize OpenCL.

Definition at line 529 of file COM_WorkScheduler.cc.

References blender::compositor::COM_is_opencl_enabled(), blender::compositor::COM_threading_model(), blender::compositor::opencl_initialize(), blender::compositor::Queue, blender::compositor::SingleThreaded, blender::compositor::Task, and blender::compositor::threading_model_queue_initialize().

Referenced by COM_execute().

◆ schedule()

void blender::compositor::WorkScheduler::schedule ( WorkPackage * package )

static

schedule a chunk of a group to be calculated. An execution group schedules a chunk in the WorkScheduler when ExecutionGroup.get_flags().open_cl is set the work will be handled by a OpenCLDevice otherwise the work is scheduled for an CPUDevice