Parallel IC Optimization via Independent Cell Instances
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current optimization techniques for integrated circuit design are inefficient in exploring a large number of possible optimizations within a reasonable time frame, especially when detailed timing analysis is required, due to high computational resource usage, leading to conservative and less accurate methods.
Innovation Solution
A parallel optimization method is employed, where independent cell instances are analyzed and replaced in parallel, ensuring that no two instances are in the same fan-in and fan-out cones, allowing for detailed timing analysis and exploring multiple alternatives efficiently, while maintaining accurate results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complete timing analysis is performed on each design change during optimization, then measurement precision is improved, but use of energy by stationary object increases
Solution Approach 1:
The patent segments the design into independent cell instances and their associated fan-in/fan-out cones. By identifying and analyzing only the specific portions of the design affected by each cell replacement (the cones), rather than performing complete timing analysis on the entire design, the method reduces computational resources while maintaining timing analysis accuracy for the affected regions.
2Measurement precision
If complete timing analysis is performed on each design change during optimization, then measurement precision is improved, but productivity decreases
Solution Approach 1:
The patent segments the design into independent cell instances and their associated fan-in/fan-out cones. By identifying and analyzing only the specific portions of the design affected by each cell replacement (the cones), rather than performing complete timing analysis on the entire design, the method reduces computational resources while maintaining timing analysis accuracy for the affected regions.
3Use of energy by stationary object
If conservative optimization techniques are used to reduce computational resources, then use of energy by stationary object decreases, but measurement precision worsens
Solution Approach 1:
The patent segments the design into independent cell instances and their associated fan-in/fan-out cones. By identifying and analyzing only the specific portions of the design affected by each cell replacement (the cones), rather than performing complete timing analysis on the entire design, the method reduces computational resources while maintaining timing analysis accuracy for the affected regions.
4Productivity
If parallel analysis of multiple cell instances is performed, then productivity improves, but measurement precision may worsen due to increased complexity
Solution Approach 1:
The patent segments the design into independent cell instances and their associated fan-in/fan-out cones. By ensuring that selected cell instances have no overlapping cones, the method creates independent analysis units that can be processed in parallel without interfering with each other, thus maintaining timing analysis accuracy while improving productivity.
Solution Approach 2:
The patent performs preliminary selection of cell instances whose fan-in and fan-out cones are mutually independent before initiating parallel timing analysis. This preliminary action ensures that the subsequent parallel analysis operations will not have overlapping computational requirements, allowing accurate parallel processing without requiring conservative sequential analysis.
Data Source
AI summary
The present invention provides a method for parallel optimization of an integrated circuit design based on the use of sets of cell instances that are independent from each other. Multiple changes to a design are analyzed in parallel by ensuring that no two cell instances that are being changed are in the same fan-in and fan-out cones. This property allows full timing analysis to be performed on a design such that multiple alternatives are explored in parallel and accurate results are obtained. By ordering the choice of cell instances to change and by ordering the alternatives to try, a greater degree of optimization is found earlier in the process.