|
| | Hardcoded (int minTasksToFuse, int maxTasksToFuse, int deviceForFusedTasks, int maxNestedConcurrency, FusionImplementation fusionImplementation) |
| | Construct hardcoded scheme.
|
| |
| virtual | ~Hardcoded ()=default |
| |
| virtual void | startBSPSection (int nestedParallelismLevel) override |
| | Notifies the strategy that we enter a BSP section.
|
| |
| virtual void | endBSPSection (int nestedParallelismLevel) override |
| | Notifies the strategy that we leave a BSP (fork-join) section.
|
| |
| virtual FuseInstruction | fuse (int taskType) override |
| | How many tasks to fuse and to which device to deploy.
|
| |
| virtual ExecutionPolicy | paralleliseForkJoinSection (int nestedParallelismLevel, int numberOfTasks, int taskType) override |
| | Determine how to handle/realise parallelisation within fork/join region.
|
| |
| virtual void | updateFuseStatistics (int taskType, int queueSize, double averageNumberOfPendingReadyTasksPerProcessTaskExecution, double probabilityThatTaskIsCompleted) override |
| | Update task fusion status.
|
| |
| virtual FusionImplementation | getFusionImplementation () const override |
| | Determine how the implementation handles fusable tasks.
|
| |
| virtual | ~Strategy ()=default |
| |
A hard coded strategy how to handle tasks and task fusion.
This hardcoded variant can be used to realise a few standard tasking patterns. As trivial example, it allows us to switch off all tasking, to map all tasks immediately onto one GPU device, or to fix the number of tasks that are to be fused.
Definition at line 25 of file Hardcoded.h.
| virtual ExecutionPolicy tarch::multicore::orchestration::Hardcoded::paralleliseForkJoinSection |
( |
int | nestedParallelismLevel, |
|
|
int | numberOfTasks, |
|
|
int | taskType ) |
|
overridevirtual |
Determine how to handle/realise parallelisation within fork/join region.
Peano models its execution with multiple parallel, nested fork/join sections. You could also think of these as mini-BSP sections. This routine guides the orchestration how to map those BSP sections onto tasks.
The decision can be guided by basically arbitrary contextual factors. The most important one for me is the nesting factor. As we work mainly with OpenMP, where tasks are tied to one core, it makes limited sense to have nested parallel fors. Notably, it makes stuff slower. So usually, I return ExecutionPolicy::RunSerially with anything with a nesting level greater than 1.
- Parameters
-
| nestedParallelismLevel | Please compare with tarch::multicore::spawnAndWait() which ensures that this flag equals 1 on the top level. A parameter of 0 would mean that no fork/join region has been opened. For such a parameter, the code would not query this function. |
| taskType | If we enter a fork-join section, this section logically spawns a set of tasks, which are all of the same type. So the task type here is given implicitly by the code location. But each BSP section has a unique identifier. |
Implements tarch::multicore::orchestration::Strategy.
| virtual void tarch::multicore::orchestration::Hardcoded::updateFuseStatistics |
( |
int | taskType, |
|
|
int | queueSize, |
|
|
double | averageNumberOfReadyTasksPerProcessTaskExecution, |
|
|
double | probabilityThatTaskIsCompleted ) |
|
overridevirtual |
Update task fusion status.
Feed orchestration with some stats around the task queue of interest. This routine is not called every time a process read tasks kicks off, but only every now and then. Furthermore, please note that this routine is invoked iff the orchestration beforehand had decided to fuse tasks.
- Parameters
-
| taskType | Number of the task that we are studying |
| queueSize | Size of the queue at this observation point |
| averageNumberOfReadyTasksPerProcessTaskExecution | average over number of tasks in the queue whenever a process task has kicked off and the underlying real task has not completed yet. |
| probabilityThatTaskIsCompleted | Probability that task result has been there already when the corresponding instance of ProcessReadyTask fires up. This is a biased value: If we try to look up if a task is already completed and the underlying dictionary is locked, then this counts as a "no hit" as well. |
Implements tarch::multicore::orchestration::Strategy.
Public Member Functions inherited from tarch::multicore::orchestration::Strategy
1.10.0