Hardware Performance Monitor Averaging for Process Variation Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Hardware performance monitors in digital circuits are affected by local process variations, leading to inaccuracies and increased power consumption and chip area due to mismatch between monitor and design performance values, which existing single-instance monitors cannot effectively compensate for.
Innovation Solution
Implementing multiple identical hardware performance monitors to measure performance values, calculating an average value to approximate ideal performance, and using this average to regulate global process variations, thereby reducing the impact of local variations and minimizing additional hardware and energy overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single hardware performance monitor is used, then device complexity is reduced, but measurement precision deteriorates due to local process variations causing mismatch between monitor and design performance values
Solution Approach 1:
The patent combines multiple hardware performance monitor instances (n≥1) into a unified monitoring system that measures performance values and calculates their average. This merging approach reduces the impact of local process variations on measurement accuracy while maintaining manageable system complexity through centralized control logic.
Solution Approach 2:
The patent changes the monitoring parameter from a single performance value to an average of multiple performance values. By measuring n performance values and computing their average, the system reduces the effect of local process variations and achieves more accurate representation of ideal performance without proportionally increasing device complexity.
2Manufacturing precision
If the active device area of the monitor circuit is increased, then manufacturing precision improves by reducing local process variation effects, but area consumption increases
Solution Approach 1:
The patent segments the monitoring function into multiple independent but identical monitor instances distributed across the chip. Instead of concentrating all monitoring resources in one large circuit, the system divides the monitoring task among n smaller instances, each less susceptible to local process variations, thereby improving manufacturing precision without proportionally increasing total area.
Solution Approach 2:
The patent changes the approach from increasing individual monitor size to increasing the number of monitor instances. By using n identical monitors with smaller individual areas and computing the average of their readings, the system achieves better robustness against local process variations while controlling overall area consumption.
3Measurement precision
If multiple identical hardware performance monitors are implemented, then measurement precision improves by reducing local process variation effects, but power consumption increases
Solution Approach 1:
The patent segments the monitoring function into multiple identical instances that can operate independently. Each instance consumes less power than a single comprehensive monitor would require, and the distributed architecture allows for potential power management strategies where not all instances need to operate at full power simultaneously, reducing overall power consumption while maintaining measurement precision.
Solution Approach 2:
The patent changes the monitoring strategy from using one high-power monitor to multiple lower-power monitors whose average reading provides accurate performance measurement. This parameter change in the monitoring architecture reduces the total power consumption while improving robustness against local process variations through statistical averaging.
Data Source
AI summary
A method and an apparatus for reducing an effect of local process variations of a digital circuit on a hardware performance monitor includes measuring a set of performance values (c1, c2 . . . cn) of the digital circuit by n identical hardware performance monitors, where n is a natural number greater than 1, determining an average value cmean of the measured performance values (c1, c2 . . . cn), as an approximation of an ideal performance value c0, selecting one performance value cj of the set of performance values (c1, c2 . . . cn) by a controller, comparing the performance value cj with a reference value cref by a controller the controller, resulting in a deviation value Δc, and controlling an actuator by using the deviation Δc for regulating the local global process variations to the approximation cmean of the ideal performance value c0.