| 摘要 |
A method and system are disclosed for generating a static partition scheduling timeline, which minimizes context switching overhead. An example method for generating a static partition scheduling timeline is disclosed, which models each partition as a period and duration, and allocates each partition its requested duration of execution time in each period. More precisely, the CPU execution time is divided into a repeating sequence of activation frames. An activation frame is defined as a continuous period of time during which no partition period starts or ends. An "earliest deadline, first prioritization" technique can be used for allocating partition execution time slices to each of the activation frames, in order to create an initial timeline. Alternatively, another suitable technique for generating an initial timeline may be used, such as, for example, a rate monotonic scheduling technique. In any event, the disclosed method allocates slices within an activation frame in a suitable order. Thus, partitions that have slices in adjacent activation frames within the same partition period are moved to the start and end of the respective activation frames and merged at the frame boundary. Specifically, a list of partitions and their respective attributes (e.g., period and duration) are provided as inputs for the method, and a static partition scheduling timeline is generated as a list of slices that repeats at the least common multiple of all of the partition periods. Each such slice is defined by a slice start time, slice duration, and the partition that the slice is allocated to. The generated scheduling timeline also includes one or more idle slices (e.g., time not allocated to a partition) so that the sum of the slice durations is equal to the period at which the timeline repeats.
|
