Processor MEP Control Using Sensor-Driven Heuristic Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current techniques fail to efficiently identify and track the minimum energy point (MEP) in digital integrated circuits, leading to suboptimal energy consumption due to their reliance on recursive and computationally expensive methods for dynamic voltage and frequency scaling.
Innovation Solution
A sensor-driven MEP control circuit that leverages heuristic techniques and pre-characterized lookup tables to quickly determine and adjust the MEP in real-time, enabling fast and efficient tracking of optimal energy consumption across varying workload and temperature conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If recursive and computationally expensive methods are used for dynamic voltage and frequency scaling, then the system can identify MEP, but the computational overhead is high and tracking speed is slow
Solution Approach 1:
The patent pre-characterizes the processor across process, voltage, and temperature (PVT) corners to build lookup tables containing MEP information before runtime. This preliminary characterization eliminates the need for recursive computation during operation, as the system directly queries pre-computed values based on current PVT conditions, thereby resolving the contradiction between accurate MEP identification and computational overhead.
Solution Approach 2:
The patent creates simplified models and lookup tables that copy essential MEP characteristics from extensive PVT simulations. Instead of performing full recursive computations, the system uses these copied representations to quickly determine operating points, maintaining accuracy while dramatically reducing computational complexity and overhead.
2Measurement precision
If conventional sweep-based energy computing methods are used, then the system can determine MEP, but the tracking speed is slow and real-time adaptation is poor
Solution Approach 1:
The patent performs comprehensive energy computations and MEP determinations during the offline PVT characterization phase, storing results in lookup tables. At runtime, the system simply queries these pre-computed values based on current PVT sensors, achieving both accurate MEP determination and fast tracking speed without performing recursive sweeps during operation.
Solution Approach 2:
The patent replaces the mechanical sweep-based exploration method with a direct lookup approach using pre-characterized data. Instead of physically sweeping through voltage and frequency points to find MEP, the system substitutes this with memory-based retrieval of pre-computed optimal points, dramatically increasing tracking speed while maintaining accuracy.
3Ease of operation
If the system operates at nominal supply voltage, then the operation is simple, but energy efficiency is suboptimal compared to near-threshold voltage operation
Solution Approach 1:
The patent implements dynamic voltage and frequency scaling that automatically adjusts operating points based on real-time PVT conditions and workload requirements. The system transitions from fixed nominal operation to adaptive dynamic operation, selecting from pre-characterized MEPs the optimal combination of voltage and frequency that maximizes energy efficiency while maintaining performance, thereby resolving the contradiction between operational simplicity and energy efficiency.
Data Source
AI summary
In one embodiment, a processor includes a minimum energy point (MEP) controller to: generate a change in thermal tracking information, based at least in part on prior and current thermal information; generate a change in activity tracking information, based at least in part on prior activity information and current activity information; and determine a MEP performance state based at least in part on the change in thermal tracking information and the change in activity tracking information. Other embodiments are described and claimed.