SoC Performance Control Using Function Execution History
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing system-on-chip (SoC) technologies face inaccuracies in performance prediction and control due to the neglect of software factors, leading to inefficient power/energy consumption and processing inefficiencies.
Innovation Solution
SoCs equipped with a history storage and performance controller that generate optimized control information based on the history of processor utilization, allowing for accurate performance prediction and adaptive control of processor performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If performance control is based only on hardware factors (current run-state), then the control system is simple, but the performance prediction accuracy deteriorates due to not reflecting software factor changes
Solution Approach 1:
The system performs preliminary action by accumulating and storing performance history data (control information and utilization information) during previous function executions. This historical data is stored in a history storage unit and later retrieved to generate optimized control information, enabling accurate performance prediction without requiring complex real-time analysis of software factors.
Solution Approach 2:
The patent introduces an intermediary mechanism (performance history information) that mediates between hardware factors and software factor changes. The history storage unit acts as an intermediary that captures the combined effects of hardware and software factors during previous executions, allowing the control system to indirectly account for software changes without directly monitoring them.
2Measurement precision
If performance control is based on comprehensive hardware and software factors, then performance prediction accuracy improves, but energy consumption increases due to unnecessary control operations
Solution Approach 1:
The system implements feedback by comparing predicted performance (based on history) with actual performance outcomes. The performance management unit uses this feedback to continuously optimize control decisions, generating optimized control information that reduces unnecessary control operations while maintaining accurate performance prediction, thereby lowering energy consumption.
Solution Approach 2:
The patent changes parameters by transitioning from using only current run-state parameters to using historical performance parameters (accumulated control information and utilization information). This parameter change enables the system to predict performance more accurately and make better-informed control decisions that reduce unnecessary energy consumption.
3Measurement precision
If performance control is based on comprehensive hardware and software factors, then performance prediction accuracy improves, but processing efficiency deteriorates due to complex analysis requirements
Solution Approach 1:
The system performs preliminary action by accumulating performance history data during function executions. This historical data is stored and organized in advance, so when performance prediction is needed, the system can quickly retrieve and process the pre-accumulated data without performing complex real-time analysis, thus maintaining high processing efficiency while achieving accurate predictions.
Solution Approach 2:
The performance history information serves the system itself by providing the necessary data for accurate performance prediction. The accumulated history data automatically reflects software factor changes and hardware performance characteristics, enabling the control system to make accurate predictions without requiring external complex analysis or additional processing resources.
Data Source
AI summary
Disclosed are system-on-chip devices (SoCs) and operating methods thereof. A system-on-chip (SOC) may include a memory including a history storage, a processor configured to execute an application, performance optimization module configured to generate and transmit optimized control information for the processor executing a function included in the application, based on a history corresponding to the function, when the processor executes the function and the history corresponding to the function is in the history storage, and a performance management module configured to generate a control signal for controlling performance of the processor, based on the optimized control information received from the performance optimization module, wherein the history includes control information for the processor and utilization information accumulated according to the control information, the control information having been accumulated during at least one time of previous execution of the function.