Semiconductor Layout Generator Assignment Using CNN Inference

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The process of assigning generators to sub-cells in semiconductor layouts is time-consuming and labor-intensive due to the complexity of hierarchical structures and increasing types of sub-cells and generator libraries, leading to decreased productivity.

Innovation Solution

A method using a computing device to divide semiconductor layouts into channels, convert them into matrices, and infer appropriate generators using a trained convolutional neural network, which includes a feature extraction block and fully-connected layer, to streamline the assignment process.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If manual generator assignment is performed for each sub-cell, then design flexibility and control are maintained, but time consumption and labor intensity increase sharply

Engineering Contradiction:
Improvedesign controlVSAvoidgenerator assignment time
Core Design Contradiction:
Ease of operationVSLoss of time

Solution Approach 1:

The system enables self-service automation where the generator assignment process performs itself through neural network inference. The CNN automatically analyzes layout images, identifies sub-cells, and assigns appropriate generators without human intervention, while designers retain oversight through the user interface for reviewing and adjusting assignments.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent replaces the manual mechanical process of generator assignment with an automated neural network system. The CNN model substitutes human designers' visual inspection and decision-making with automated image recognition and pattern matching, dramatically reducing time consumption while maintaining assignment quality.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Adaptability or versatility

If the number of sub-cell types and generator library types increases, then design versatility improves, but the complexity of generator assignment increases

Engineering Contradiction:
Improvedesign versatilityVSAvoidassignment complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The neural network system serves as a universal platform that handles diverse sub-cell types and generator libraries through a single unified interface. The CNN model learns to recognize multiple sub-cell patterns and map them to appropriate generators from various libraries, eliminating the need for separate assignment processes for each cell type.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system manages complexity by transforming the assignment problem into parameter-based image recognition. Instead of dealing with complex hierarchical names and specifications, the CNN converts visual layout features into parameter representations that directly guide generator selection, simplifying the mapping between diverse cell types and generators.

Inventive Principle:
Principle #35Parameter changes

3Productivity

If automated generator assignment is implemented, then productivity increases, but the requirement for trained neural network models and computational resources increases

Engineering Contradiction:
Improvelayout design productivityVSAvoidsystem complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The system performs preliminary action by pre-training the neural network model on comprehensive datasets of sub-cells and generators before actual design work. This upfront training phase creates a ready-to-use assignment system that can rapidly process layouts without requiring complex real-time computations during the actual design process.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent introduces an intermediary user interface layer that mediates between the automated neural network system and the designer. This interface simplifies the complex automated process into intuitive interactions, allowing designers to initiate assignments, review results, and make adjustments without needing to understand the underlying neural network complexity.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP4708115A1Generator allocation method for semiconductor layout, allocation training method, and operation device
Publication Date: 2026.03.11 POSTECH ACADEMY INDUSTRY FOUNDATION
  • EP4708115A1 patent drawingFigure 1
  • EP4708115A1 patent drawingFigure 2
  • EP4708115A1 patent drawingFigure 3

AI summary

Provided are a computing device and method for assigning a generator to a semiconductor layout and a method of training a neural network. The former method includes an input operation of receiving a layout by a computing device, a division operation of dividing the layout into a plurality of channels, a conversion operation of converting each of the divided channels into a matrix, and an inference operation of inferring a generator to be assigned to the layout from the matrix. The inference operation is performed by displaying one or more generator candidates corresponding to the received layout.