Multi-core Processor Power Control via Local Energy Counters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In multi-core processors, it is challenging to control power consumption effectively due to the difficulty in measuring and managing power usage across processor cores, especially in systems with shared resources like caches, which can lead to excessive power consumption and malfunction.
Innovation Solution
Each processor core calculates consumed energy based on issued instruction information and power consumption information, allowing for the restriction of new instruction issuance to manage power usage and set upper limits on power consumption, thereby preventing excessive energy use.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a huge information collection network is implemented to measure power consumption directly in a multi-core processor with shared cache, then measurement precision of power consumption is improved, but device complexity and manufacturing cost increase significantly
Solution Approach 1:
The patent extracts the power consumption measurement function from the complex information collection network and implements it within each processor core using local counters. Each core independently counts its own power consumption events without requiring external monitoring infrastructure, thereby eliminating the need for a huge information collection network while maintaining measurement precision.
Solution Approach 2:
Each processor core performs self-measurement of its power consumption using local counters that count power consumption events generated by that core. This self-service approach eliminates the need for external monitoring systems and complex information collection networks, as each core independently tracks and reports its own power usage.
2Measurement precision
If power consumption is controlled for each processor core individually, then power consumption control precision is improved, but ease of operation deteriorates due to difficulty in measuring and managing power usage across cores
Solution Approach 1:
The patent segments the power consumption management system into independent counter units distributed across each processor core. Each core has its own counter that independently tracks power consumption events, enabling precise per-core control without requiring complex centralized management. This segmentation simplifies operation by making each core self-sufficient in power monitoring.
Solution Approach 2:
The patent introduces counter units as intermediary components within each processor core that mediate between power consumption events and control logic. These counters act as simple intermediaries that convert physical power consumption events into countable data, facilitating easy management and control of power usage across multiple cores without direct complex interactions.
3Productivity
If all transistors are caused to operate simultaneously to maximize processing capability, then productivity is improved, but use of energy increases excessively leading to the dark silicon problem
Solution Approach 1:
The patent implements dynamic power management by enabling processor cores to adaptively control their operation based on real-time power consumption conditions. Each core monitors its own power usage through local counters and dynamically adjusts its activity level, allowing the system to maximize productivity when power is available while preventing excessive energy consumption when power limits are approached, thus solving the dark silicon problem.
Data Source
AI summary
In a multi-core processor having a plurality of processor cores, each of the processor cores is configured to calculate consumed energy on a basis of issued instruction information and power consumption information. The consumed energy is energy consumed in execution of an instruction issued by the processor core. The issued instruction information represents the instruction issued by the processor core. The power consumption information represents change of power consumption resulting from execution of processing corresponding to the instruction.