Double Clustering for Semiconductor Device Placement Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The current semiconductor design process is complex and time-consuming, relying heavily on engineer experience, making it difficult to efficiently place tens to millions of semiconductor devices while maintaining connection relationships, leading to inconsistent design quality and high costs.
Innovation Solution
A method using double clustering, where semiconductor devices are grouped into clusters based on connection relationships, with further sub-clustering within these clusters, utilizing netlist information and hierarchy data to evaluate and optimize placement, incorporating reinforcement learning and electronic design automation for simulation and partitioning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If engineer experience and intuition are used for semiconductor device placement, then design flexibility is maintained, but design consistency deteriorates and time consumption increases
Solution Approach 1:
The patent segments the semiconductor device placement task into multiple evaluation stages: initial placement evaluation, cluster-based grouping evaluation, and detailed connection relationship evaluation. This segmentation allows automated processing of routine evaluation tasks while preserving engineer expertise for critical decision points, thereby reducing time consumption without sacrificing design flexibility.
Solution Approach 2:
The patent introduces an automated evaluation system as an intermediary between the engineer's intuition and the final placement design. This intermediary systematically evaluates placement candidates based on multiple criteria (connection relationships, device types, density), providing consistent recommendations that reduce time consumption while maintaining the adaptability of engineer-driven design processes.
2Manufacturing precision
If tens to millions of semiconductor devices are evaluated for placement, then placement accuracy is improved, but evaluation complexity increases
Solution Approach 1:
The patent divides the evaluation of tens to millions of devices into hierarchical levels: first evaluating devices in clusters based on connection relationships, then evaluating individual devices within each cluster. This segmentation reduces evaluation complexity by processing devices in manageable groups while maintaining placement accuracy through multi-level verification.
Solution Approach 2:
The patent adds a cluster-based dimensional layer to the traditional device-level evaluation. By organizing devices into clusters based on connection relationships and evaluating at this intermediate dimension, the system manages the complexity of millions of devices while preserving placement accuracy through the additional organizational structure.
3Manufacturing precision
If connection relationships of all semiconductor devices are considered, then placement quality is improved, but evaluation time increases
Solution Approach 1:
The patent segments the connection relationship evaluation into cluster-level and device-level assessments. By first evaluating connection relationships at the cluster level and then refining at the individual device level, the system maintains placement quality while reducing evaluation time through hierarchical processing.
Solution Approach 2:
The patent performs preliminary evaluation of connection relationships at the cluster level before conducting detailed device-level assessment. This preliminary action identifies promising placement regions and connection patterns early, allowing the system to maintain high placement quality while significantly reducing the time required for comprehensive evaluation of all devices.
Data Source
AI summary
Disclosed is a method of performing double clustering to evaluate placement of semiconductor devices performed by a computing device according to an exemplary embodiment of the present disclosure. The method includes receiving connection relationship information representing a connection relationship between semiconductor devices, perform clustering on the semiconductor devices by utilizing first reference information based on the connection relationship information, and perform sub-clustering in a cluster generated by the clustering, by utilizing second reference information based on the connection relationship information.