Multi-Processor Task Scheduling for Power and Performance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multi-core processors face challenges in managing tasks related to user interaction, leading to schedule contention, increased power consumption, and decreased operating performance due to inefficient task boosting and scheduling.
Innovation Solution
An electronic device with at least one high-performance processor and one low-performance processor, where tasks are assigned to boosting and non-boosting foreground control groups based on user input and boot completion, with the scheduler managing task distribution between the processors to optimize performance and power usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If all foreground tasks are assigned to the boosting foreground control group and boosted to provide user interaction, then user interaction performance is improved, but schedule delay of applications occurs and unnecessary power consumption increases
Solution Approach 1:
The patent divides foreground tasks into two separate control groups: boosting foreground control group and non-boosting foreground control group. This segmentation allows the system to selectively boost only specific tasks that require rapid user interaction response while leaving other tasks unboosted, thereby avoiding unnecessary power consumption from boosting all tasks uniformly.
Solution Approach 2:
The patent applies different quality levels of processing to different tasks by assigning them to different control groups. Tasks in the boosting group receive high-priority scheduling and CPU frequency boosting (local high quality), while tasks in the non-boosting group receive standard processing (local standard quality), optimizing overall system efficiency rather than uniformly applying high performance to all tasks.
2Speed
If tasks are migrated to a big cluster (high-performance core) for user interaction, then user interaction performance is improved, but schedule contention increases and power consumption increases
Solution Approach 1:
The patent segments tasks into boosting and non-boosting foreground control groups, allowing high-priority user interaction tasks to be migrated to the big cluster for fast processing while keeping other tasks on the little cluster. This prevents schedule contention on the big cluster by avoiding unnecessary migration of tasks that do not require high-performance processing.
Solution Approach 2:
The patent applies boosting action partially rather than excessively - only to tasks that genuinely require rapid user interaction response. By selectively boosting specific tasks rather than all foreground tasks, the system achieves necessary user interaction performance without causing schedule contention and productivity loss from over-utilizing the big cluster.
3Speed
If the CPU clock is increased to process user interaction tasks, then user interaction performance is improved, but power consumption increases
Solution Approach 1:
The patent applies CPU frequency boosting locally and selectively only to tasks in the boosting foreground control group that require rapid user interaction response. This localized quality enhancement avoids the energy waste of increasing CPU clock for all tasks uniformly, optimizing the balance between processing speed and power consumption.
Solution Approach 2:
The patent applies frequency boosting action partially rather than excessively - only to specific user interaction tasks that require fast response. This partial application of boosting avoids unnecessary power consumption while maintaining adequate performance for tasks that do not require high-speed processing.
Data Source
AI summary
An electronic device includes at least one processor including a first processor and a second processor separate from the first processor, a memory electrically connected to the at least one processor and storing instructions, wherein the at least one processor is further configure to execute the instructions to assign foreground tasks to a boosting foreground control group and a non-boosting foreground control group in response to a user input, based on completion of booting of the electronic device, schedule at least one task assigned to the boosting foreground control group for the first processor, and schedule at least one task assigned to the non-boosting foreground control group for the second processor, and performance of the second processor may be lower than performance of the first processor.