Dynamic Current Modeling via Adaptive Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing dynamic current prediction models for integrated circuits require enormous computational resources, making them inefficient for simulating power consumption and power grid noise in modern IC designs with millions of gates.
Innovation Solution
The development of a predictor that generates accurate dynamic current analyses by using a large set of training data from combinations of transition slew times and fanout models, and employs adaptive clustering to reduce computational resources without compromising accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If detailed dynamic current prediction models are used for accurate analysis, then measurement precision is improved, but use of energy and computational resources increase enormously
Solution Approach 1:
The circuit design is divided into clusters of similar instances (gates, buffers, etc.) based on their electrical characteristics. Each cluster is represented by a representative instance, reducing the number of detailed simulations needed while maintaining analysis accuracy for the entire design.
Solution Approach 2:
Adaptive clustering selectively applies detailed analysis only to representative instances of each cluster rather than every individual instance. This partial action approach maintains measurement precision for critical elements while significantly reducing overall computational resources required.
2Measurement precision
If detailed dynamic current prediction models are used for accurate analysis, then measurement precision is improved, but productivity decreases due to enormous computational requirements
Solution Approach 1:
The design is segmented into clusters of electrically similar instances. By analyzing only representative instances from each cluster rather than every individual instance, the simulation achieves detailed accuracy where needed while dramatically improving overall productivity through reduced computational scope.
Solution Approach 2:
Instances are pre-clustered based on their electrical characteristics before detailed dynamic current analysis is performed. This preliminary grouping action enables the subsequent simulation to focus computational resources on representative instances only, improving both accuracy and productivity.
3Use of energy by moving object
If adaptive clustering is applied to reduce computational resources, then use of energy is reduced, but measurement precision may be compromised
Solution Approach 1:
The clustering approach is adaptive rather than static, adjusting the level of detail and cluster representation based on the specific electrical characteristics and importance of different circuit elements. This dynamic adaptation maintains measurement precision for critical instances while reducing computational resources for less critical ones.
Solution Approach 2:
The clustering algorithm changes parameters such as cluster size, representative instance selection, and analysis depth based on the electrical characteristics of each cluster. This parameter adaptation ensures that measurement precision is maintained for electrically significant instances while reducing computational resources overall.
4Productivity
If adaptive clustering is used to improve productivity, then simulation speed increases, but measurement precision may be reduced
Solution Approach 1:
The design is segmented into clusters with representative instances that capture the essential electrical behavior. This segmentation enables high-productivity simulation by analyzing fewer instances while maintaining current waveform accuracy through careful selection of representative examples from each cluster.
Solution Approach 2:
Representative instances are created as copies or models that represent entire clusters of similar circuit elements. These copied representative instances are analyzed in detail, and their results are applied to all members of the cluster, improving productivity while maintaining measurement precision through accurate representation.
Data Source
AI summary
Circuit design techniques can use a trained predictor to predict key dynamic current metrics (such as peak current, peak time, pulse width and total charge) for a gate in a circuit library, where the predictor has been trained over different combinations of different input transition slews and different output fanout models. A dynamic current model solver can be used for a gate in the cell library to derive waveforms (of current versus time) for the different combinations, and a predictor, such as a neural network, can be trained with the outputs from the solver for the different combinations. The trained predictor can be used in a runtime simulation to solve for the dynamic current demand model of the various gates in a circuit design (such as all of the gates in an integrated circuit).