Public Member Functions
|MultithreadScheduler (int prefetchRate=0, int prefetchQueue=0, float frameRate=0.0f, int nThreads=0)|
|Creates a new multithread scheduler. |
|Deletes this scheduler. |
|virtual bool||supportsPrefetch (bool gpuTasks)|
|Returns true if the prefetch rate or the fixed frame rate is not null, or if there are several threads and gpuTasks is false. |
|virtual void||schedule (ptr< Task > task)|
|Adds a task whose deadline is not immediate. |
|virtual void||reschedule (ptr< Task > task, Task::reason r, unsigned int deadline)|
|Forces the reexecution of the given task and of its sub tasks. |
|virtual void||run (ptr< Task > task)|
|Executes the given tasks. |
|void||monitorTask (const string &taskType)|
|Adds the given task type to the tasks whose execution times must be monitored (debug). |
Protected Member Functions
|void||init (int prefetchRate, int prefetchQueue, float frameRate, int nThreads)|
|Initializes this scheduler. |
A Scheduler that can use multiple threads.
This scheduler can work with one or more threads, and it can try to follow a fixed framerate (i.e. a number of calls to run per second). If a prefetch rate is specified, the main thread executes at least the specified number of prefetching tasks par frame (when such tasks are available), after all the tasks for the current frame have been executed. Hence if a prefetch rate is specified, or if a fixed frame rate is specified, this scheduler supports prefetching of tasks of any kind. Otherwise, if several threads are used, prefetching of cpu tasks is supported, but not prefetching of gpu tasks.
|ork::MultithreadScheduler::MultithreadScheduler||(||int|| prefetchRate =
|int|| prefetchQueue =
|float|| frameRate =
|int|| nThreads = |
Creates a new multithread scheduler.
|prefetchRate||the minimum number of prefetch task to execute at each frame, after all the tasks for the current frame have been executed (and if prefetching tasks are available). The prefetching of gpu tasks is only possible if this rate is not 0.|
|prefetchQueue||the maximum number of prefetching tasks that can be queued for execution. If a prefetch rate or a fixed frame rate is specified, this value must not be 0. When the queue is full, new prefetch tasks are simply discarded and schedule returns false. This maximum queue size prevents the number of prefetching tasks to grow unbounded, if new prefetching tasks are generated at a greater rate than the rate at which they are executed.|
|frameRate||a fixed framerate that this scheduler should try to follow, or 0 to not fix any framerate.|
|nThreads||the number of threads to use in addition to the main thread of the application. Hence 0 means that only one thread will be used, the main application thread.|
Deletes this scheduler.
Initializes this scheduler.
|void ork::MultithreadScheduler::monitorTask||(||const string &||taskType||)|
Adds the given task type to the tasks whose execution times must be monitored (debug).
|virtual void ork::MultithreadScheduler::reschedule||(||ptr< Task >||task,|
Forces the reexecution of the given task and of its sub tasks.
|task||a task or task graph that must be reexecuted. This task is marked as undone (with Task::setIsDone) so that it will be reexecuted.|
|r||the reason why the task must be reexecuted.|
|deadline||the frame number before which this task must be reexecuted.|
Executes the given tasks.
This method does not return before all tasks with an immediate deadline are completed.
|task||a task or task graph to be executed.|
Adds a task whose deadline is not immediate.
This method must not be called if this scheduler does not support prefetch (see supportsPrefetch). Otherwise it adds this task and its sub tasks to the list of tasks to be executed by this scheduler, and returns immediately (i.e. before these tasks are executed).
|task||a task or task graph whose deadline is not immediate.|
|virtual bool ork::MultithreadScheduler::supportsPrefetch||(||bool||gpuTasks||)||
Returns true if the prefetch rate or the fixed frame rate is not null, or if there are several threads and gpuTasks is false.
In addition, if there is only one thread for cpu tasks, and for gpu tasks in all cases, the curent number of prefetching tasks waiting for execution must be less than the prefetch queue size, otherwise this method returns false.