Dynamic Workload Switching Between Heterogeneous Processor Cores
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As performance demands increase in computing devices, multiple core processors face challenges in balancing performance and power consumption, particularly in portable platforms where total available power is limited, and existing mechanisms for dynamic scaling are complex and inefficient.
Innovation Solution
A prediction algorithm dynamically switches workloads between cores with different characteristics, using dynamic workload characterization and scheduling information to select appropriate cores for execution, allowing for transparent and efficient performance adjustments based on predicted performance needs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple core processors are used to increase performance, then processing capability is improved, but power consumption increases
Solution Approach 1:
The system dynamically switches between different core types (high-performance and low-power cores) based on workload characteristics and performance requirements. This dynamic adaptation allows the processor to optimize the balance between processing capability and power consumption by selecting the appropriate core type for each task, rather than statically allocating all tasks to high-performance cores.
Solution Approach 2:
The patent applies different core types to different workloads based on their specific requirements. High-performance cores are used for computationally intensive tasks requiring maximum processing power, while low-power cores handle less demanding tasks. This local optimization ensures that each workload is executed on the most appropriate core type, minimizing overall power consumption while maintaining necessary performance levels.
2Device complexity
If fixed scheduling of workloads to particular cores is implemented, then scheduling complexity is reduced, but adaptability to performance demands decreases
Solution Approach 1:
The system employs a prediction algorithm that automatically analyzes workload characteristics and makes intelligent decisions about core selection without requiring complex manual scheduling configurations. The algorithm self-adapts to performance demands by predicting future workload requirements and proactively selecting appropriate cores, thereby maintaining low scheduling complexity while achieving high adaptability to varying performance needs.
Solution Approach 2:
The prediction algorithm performs preliminary analysis of workload characteristics before task execution begins. By predicting performance requirements in advance and pre-selecting appropriate cores, the system avoids the need for complex real-time scheduling decisions, thereby reducing scheduling complexity while maintaining high adaptability to performance demands.
3Adaptability or versatility
If dynamic scaling of performance is implemented, then adaptability to workload demands is improved, but overhead and latency increase
Solution Approach 1:
The prediction algorithm performs preliminary analysis of workload characteristics and makes core selection decisions before task execution begins. By predicting performance requirements in advance and pre-configuring the appropriate core allocation, the system minimizes the overhead and latency that would otherwise occur during runtime dynamic scaling, thereby achieving fast adaptation to workload demands.
4Use of energy by moving object
If heterogeneous cores are used to optimize power-performance balance, then energy efficiency is improved, but device complexity increases
Solution Approach 1:
The system employs an automated prediction algorithm that intelligently manages the complexity of coordinating heterogeneous cores. The algorithm automatically analyzes workload characteristics, predicts performance requirements, and makes optimal core selection decisions without requiring complex manual intervention or sophisticated scheduling software, thereby maintaining energy efficiency while managing device complexity through automation.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In one embodiment, a policy manager may receive operating system scheduling information, performance prediction information for at least one future quantum, and current processor utilization information, and determine a performance prediction for a future quantum and whether to cause a switch between asymmetric cores of a multicore processor based at least in part on this received information. Other embodiments are described and claimed.