Neural Network Toggle Estimation for IC Power Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for estimating toggle counts in integrated circuits are time-consuming and inaccurate due to the non-linearity in logic functions, making it difficult to estimate power consumption effectively.
Innovation Solution
A neural network system is trained using simulation data to estimate toggle counts by learning the nonlinear relationship between input signals and register values, allowing for more accurate and efficient prediction of toggle events.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full simulation is used to dump waveform to determine toggle count, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The patent applies preliminary action by pre-training a neural network model using full simulation data to learn the relationship between input signals and toggle counts. Once trained, the model can rapidly estimate toggle counts without performing full simulation again, thus achieving accurate measurements while reducing time loss in subsequent evaluations.
Solution Approach 2:
The patent creates a computational copy of the simulation process through a neural network model. Instead of repeatedly performing time-consuming full simulations, the trained network model serves as a lightweight copy that replicates the toggle count estimation function, enabling fast and accurate predictions without re-running complete simulations.
2Loss of time
If heuristic and regression methods are used to estimate toggle count, then loss of time is reduced, but measurement precision deteriorates
Solution Approach 1:
The patent substitutes traditional heuristic and regression methods with a neural network-based machine learning approach. This replacement enables the system to capture complex nonlinear relationships in logic functions more effectively than conventional methods, achieving both reduced estimation time and improved measurement precision through learned patterns from training data.
Solution Approach 2:
The patent changes the fundamental parameters of the estimation approach by transitioning from simple heuristic rules and linear regression to a trained neural network model. This parameter change allows the system to adapt to the specific characteristics of the circuit being analyzed, improving accuracy while maintaining fast estimation speeds through the model's efficient inference process.
3Measurement precision
If neural network system is trained to estimate toggle count, then measurement precision is improved, but use of energy increases
Solution Approach 1:
The patent performs the energy-intensive training process as a preliminary action that is executed once beforehand. The trained model is then reused for multiple subsequent estimations, amortizing the initial energy investment across numerous predictions. This approach achieves high measurement precision while reducing the average energy consumption per estimation compared to repeated full simulations.
Data Source
AI summary
A method of estimating a toggle count of a circuit, includes, in part, simulating the circuit to generate training data and an associated training toggle count of the internal nodes of the circuit in response to a test bench, training a neural network system to generate an estimate of the training toggle count in accordance with the training data and the associated training toggle count, simulating the circuit to generate simulation data in response to a first set of input values applied to the circuit, and invoking the trained neural network system to estimate a number of toggles of the internal nodes of the circuit from the simulation data. The training data may include, in part, values of input signals applied to the circuit and values of registers disposed in the circuit for a multitude of time stamps. The neural network system may include, in part, at least three layers.