Dynamic Logical Core Remapping for Thermal and Power Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multicore processors face inefficiencies due to manufacturing variations leading to differing power/performance characteristics among cores, which are typically masked by configuring all cores equally, resulting in lost opportunities for performance, power savings, thermal management, and reliability.
Innovation Solution
Implementing a controller for OS-transparent workload migration between cores, using dynamic core identifiers to remap logical cores to different physical cores based on power, performance, and thermal data, and employing integrated voltage regulators for fine-grained control of voltage and frequency of each core.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all cores are configured with the same capabilities to mask manufacturing variations, then system reliability is improved, but performance optimization and power efficiency are lost
Solution Approach 1:
The patent implements per-core capability identification and configuration, where each core is individually characterized and assigned optimal operating parameters based on its specific manufacturing variations. This allows each core to operate at its local optimum rather than forcing all cores to use a conservative uniform configuration.
Solution Approach 2:
The system dynamically remaps logical processor identifiers to physical cores based on workload characteristics and core capabilities. This dynamic remapping allows the system to adaptively assign workloads to the most suitable cores, optimizing performance while maintaining reliability through continuous monitoring and adjustment.
2Stability of the object's composition
If all cores are configured with the same capabilities to mask manufacturing variations, then system stability is improved, but power savings opportunities are lost
Solution Approach 1:
Each core is individually characterized to determine its power consumption characteristics and performance capabilities. This per-core characterization enables the system to identify and utilize cores with better power efficiency for specific workloads, reducing overall power consumption while maintaining system stability.
Solution Approach 2:
The system dynamically adjusts core assignments and operating parameters based on power management policies and real-time performance monitoring. This allows the system to optimize power consumption by assigning workloads to the most power-efficient available cores while maintaining stable operation.
3Device complexity
If all cores are configured with the same capabilities to mask manufacturing variations, then thermal management is simplified, but thermal efficiency is reduced
Solution Approach 1:
The system implements per-core thermal characterization and monitoring, allowing it to identify cores with better thermal characteristics and assign heat-sensitive workloads to those cores. This localized thermal management improves thermal efficiency while maintaining manageable complexity through systematic characterization.
Solution Approach 2:
The system dynamically remaps workloads between cores based on thermal conditions and core thermal characteristics. This dynamic thermal management allows the system to optimize heat distribution and prevent hotspots by continuously adjusting workload assignments according to real-time thermal monitoring data.
4Productivity
If dynamic remapping of logical cores to physical cores is implemented, then performance optimization is improved, but system complexity increases
Solution Approach 1:
The patent introduces a mapping layer between logical processor identifiers and physical cores that acts as an intermediary. This mapping mechanism abstracts the complexity of per-core optimization from the operating system and applications, allowing performance optimization through remapping while presenting a simplified interface to software.
Solution Approach 2:
The remapping mechanism serves multiple functions including performance optimization, power management, thermal management, and fault tolerance. By consolidating these diverse optimization goals into a single unified remapping framework, the system achieves multiple benefits while managing complexity through a multi-functional approach.
Data Source
AI summary
In one embodiment, a processor includes a plurality of cores each including a first storage to store a physical identifier for the core and a second storage to store a logical identifier associated with the core; a plurality of thermal sensors to measure a temperature at a corresponding location of the processor; and a power controller including a dynamic core identifier logic to dynamically remap a first logical identifier associated with a first core to associate the first logical identifier with a second core, based at least in part on a temperature associated with the first core, the dynamic remapping to cause a first thread to be migrated from the first core to the second core transparently to an operating system. Other embodiments are described and claimed.