DFVS Multiprocessor Power Optimization via Linear Programming
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing power management techniques for processors in mobile devices, such as Dynamic Voltage and Frequency Scaling (DVFS), are limited in handling concurrent tasks with cyclic dependencies and assume continuous frequency levels, leading to suboptimal power consumption and real-time requirement challenges.
Innovation Solution
Formulating tasks into a graph with defined dependencies and allowable discrete frequencies, solving a linear programming problem to minimize power dissipation across processors, with instructions or a local supervisor controlling operating point switches to optimize power consumption while meeting real-time requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If DVFS is used to reduce power consumption by lowering frequency and voltage, then energy consumption decreases, but processing speed and real-time requirement fulfillment deteriorate
Solution Approach 1:
The patent implements dynamic frequency and voltage scaling by allowing each processor to independently switch between multiple operating points based on real-time task requirements. The system dynamically adjusts operating parameters during runtime rather than using static configurations, enabling processors to adapt their performance and power consumption characteristics to match actual computational demands while fulfilling real-time deadlines
Solution Approach 2:
The system changes physical parameters (frequency and voltage) of processor operation by defining discrete operating points with specific frequency-voltage pairs. The linear programming formulation optimizes selection among these parameter combinations to minimize power consumption while ensuring real-time constraints are met, allowing systematic exploration of the parameter space to find optimal operating conditions
2Use of energy by moving object
If discrete operating points are used instead of continuous frequency levels, then power consumption optimization improves, but system complexity increases
Solution Approach 1:
The patent segments the continuous frequency spectrum into discrete operating points, each with predefined frequency and voltage characteristics. This segmentation simplifies the control problem by providing a finite set of selectable states rather than requiring continuous adjustment, making the system more manageable while still enabling effective power optimization through selective operation at appropriate discrete levels
Solution Approach 2:
The operating points are pre-defined and characterized before runtime execution. The system prepares a finite set of valid frequency-voltage combinations in advance, which are then selected based on real-time requirements. This preliminary characterization simplifies runtime decision-making and reduces computational complexity during actual task execution
3Use of energy by moving object
If linear programming is used to optimize power consumption, then power efficiency improves, but computation time for solving the optimization problem increases
Solution Approach 1:
The system performs static scheduling and operating point selection at compile time rather than runtime. The linear programming problem is solved in advance to determine the optimal operating point for each task segment, eliminating the need for complex real-time optimization computations. This preliminary action shifts the computational burden to offline preparation, enabling fast runtime execution
4Ease of operation
If all processors switch frequency operating points simultaneously, then synchronization is simplified, but flexibility to optimize individual processor power consumption decreases
Solution Approach 1:
The patent divides the system into independent processor units, each capable of autonomous operating point selection. Instead of forcing simultaneous frequency changes across all processors, each processor independently selects its operating point based on its specific task requirements and power constraints, enabling granular power optimization while maintaining system coordination through the centralized scheduler
Solution Approach 2:
Each processor is allowed to have different operating characteristics and frequency selections tailored to its local requirements. The system applies different operating points to different processors based on their individual task loads and power constraints, rather than enforcing uniform frequency changes across the entire system, thereby optimizing local power efficiency
Data Source
AI summary
One or more tasks to be executed on one or more processors are formulated into a graph, with dependencies between the tasks defined as edges in the graph. In the case of a Radio Access Technology (RAT) application, the graph is iterative, whereby each task may be activated a number of times that may be unknown at compile time. A discrete number of allowable frequencies for processors while executing tasks are defined, and the power dissipation of the processors at those frequencies determined. A linear programming problem is then formulated and solved, which minimizes the overall power dissipation across all processors executing all tasks, subject to several constraints that guarantee complete and proper functionality. The switching of processors executing the tasks between operating points (frequency, voltage) may be controlled by embedding instructions into the tasks at design or compile time, or by a local supervisor monitoring execution of the tasks.