Neural Network Physical Synthesis for Circuit Timing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional electronic design automation (EDA) systems face inefficiencies in applying physical synthesis optimizations for circuit designs, often resulting in wasted time and degraded performance due to the lack of effective methods to determine which optimizations will improve timing constraints in signal paths.
Innovation Solution
The use of artificial neural networks to evaluate features of signal paths not meeting timing requirements, allowing for selective application of optimizations based on predicted Quality of Result (QOR) improvements, thereby minimizing unnecessary optimizations and maximizing performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional EDA systems apply physical synthesis optimizations to all signal paths, then timing constraints may be improved, but runtime increases significantly due to unnecessary optimizations
Solution Approach 1:
The system uses machine learning models trained on historical optimization data to automatically predict which signal paths will benefit from optimizations, eliminating the need for exhaustive trial-and-error approaches. The models self-improve by learning from past results, enabling selective application of optimizations only where needed.
Solution Approach 2:
The system changes the approach from uniform optimization of all signal paths to selective optimization based on predicted effectiveness parameters. By evaluating features such as path length, logic depth, and fan-out, the system identifies candidate paths and applies optimizations only to those with high predicted improvement potential.
2Reliability
If conventional EDA systems perform exhaustive optimization attempts, then more timing issues may be resolved, but performance degrades due to wasted computational resources
Solution Approach 1:
Instead of performing exhaustive optimization on all signal paths, the system applies partial action by selectively optimizing only those paths predicted to benefit. The machine learning models identify a subset of candidate paths, and optimizations are applied only to this subset, avoiding wasted computational resources on paths unlikely to improve.
Solution Approach 2:
The system implements feedback loops where optimization results are fed back into the machine learning models for continuous improvement. Historical data from successful and unsuccessful optimization attempts are used to retrain models, improving their ability to predict effective optimization candidates and reducing wasted computational effort over time.
3Reliability
If physical synthesis optimizations are applied without selection, then timing constraints may be met, but the number of optimization attempts increases unnecessarily
Solution Approach 1:
The system performs preliminary evaluation of signal paths using machine learning models before applying optimizations. By assessing features such as path characteristics, logic element types, and congestion metrics in advance, the system identifies candidate paths that are likely to benefit from optimizations, reducing the total number of optimization attempts needed.
Data Source
AI summary
Physical synthesis for a circuit design can include determining, using a processor, features relating to a signal path of the circuit design not meeting a timing requirement, processing the features through a first neural network model using the processor, wherein the first neural network model is trained to indicate an effectiveness of a first physical synthesis optimization, and selectively performing, using the processor, the first physical synthesis optimization for the signal path based upon a result from the first neural network model.