Heterogeneous Core Scheduling via Performance Ratio Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Heterogeneous multiprocessor computing systems face challenges in scaling due to increasing power consumption and heat generation, which restricts the optimization of application scheduling across diverse processor cores for performance and power efficiency.
Innovation Solution
The implementation of a processor core modeling prediction heuristic and scheduling logic that dynamically schedules applications among heterogeneous cores based on performance and power consumption metrics, using a signature-based approach and performance counters to optimize core allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If more processors are added to improve performance, then computing performance is improved, but power consumption increases
Solution Approach 1:
The system segments processors into homogeneous groups (first set and second set) with different performance characteristics. Applications are divided and assigned to specific processor groups based on their computational requirements, allowing the system to achieve high performance for demanding tasks while keeping power consumption low by not activating all processors.
Solution Approach 2:
Different processor groups are assigned different local qualities (performance characteristics) - one group optimized for high performance and another for low power consumption. The scheduling system selectively activates appropriate processor groups based on application requirements, providing local optimization of the power-performance tradeoff.
2Productivity
If more processors are added to improve performance, then computing performance is improved, but heat generation increases
Solution Approach 1:
The system segments processors into homogeneous groups with different thermal characteristics. By assigning applications to specific processor groups based on thermal awareness, the system can achieve required performance while minimizing heat generation by not activating all processors simultaneously.
Solution Approach 2:
The system converts the harmful effect of heat generation into a useful scheduling criterion. Thermal-aware scheduling uses heat generation patterns to make intelligent decisions about processor allocation, turning thermal constraints into opportunities for optimization rather than mere limitations.
3Productivity
If application scheduling is optimized across heterogeneous cores, then performance efficiency is improved, but system complexity increases
Solution Approach 1:
The scheduling system is segmented into modular components: application performance modeling, thermal awareness modeling, and scheduling decision logic. Each component handles a specific aspect of the scheduling problem, making the overall complex system manageable through functional decomposition.
Solution Approach 2:
The system performs preliminary actions by pre-modeling application performance characteristics and thermal behavior before actual scheduling decisions are made. This advance preparation enables faster, more efficient scheduling decisions without adding complexity to the real-time scheduling process.
Data Source
AI summary
Methods and apparatus to schedule applications in heterogeneous multiprocessor computing platforms are described. In one embodiment, information regarding performance (e.g., execution performance and/or power consumption performance) of a plurality of processor cores of a processor is stored (and tracked) in counters and/or tables. Logic in the processor determines which processor core should execute an application based on the stored information. Other embodiments are also claimed and disclosed.