Parallel program task processing method and parallel program task processing device

Abstract

The invention discloses a parallel program task processing method and a parallel program task processing device. The method comprises the following steps: separately distributing a unprocessed task from current unprocessed tasks to each processing core according to a first predetermined rule; when the processing cores process the currently distributed tasks, selecting a task associated to the currently processed tasks by the processing cores from the unprocessed tasks according to data succession among the tasks when the processing cores process the currently distributed tasks; when no follow-up tasks are available, distributing an unprocessed task to the processing cores from the current unprocessed tasks according to a second predetermined rule; and after the processing cores process the currently distributed tasks, storing the obtained task data to the cache corresponding to the processing cores and processing the currently distributed tasks by taking the follow-up tasks as the processing cores till all tasks are processed. According to the method and the device disclosed by the invention, several tasks with relevance use same cache, so that the processing efficiency is high.

Description

technical field [0001] The invention relates to the technical field of process processing, in particular to a parallel program task processing method and a device thereof. Background technique [0002] In the current parallel program task processing, the commonly used method is the orderly and irregular parallel program task processing method. The parallel task processing process under this rule has three characteristics. The first is that the tasks must follow the overall or partial order (basically time sequence); the second is that the task may have data dependencies that are not known in advance; the third is that the task may dynamically create subtasks and plan to run them in the future. [0003] In this way, since multiple processing cores process in parallel, and the tasks are assigned to each processing core in sequence according to the time sequence, it is easy for several related tasks to be processed by different processing cores, which leads to subsequent tasks ...

Images