Rectilinear Block Placement via Reinforcement Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The manual placement of circuit blocks in integrated circuit design is time-consuming and often results in suboptimal solutions due to the complexity of optimizing wirelength and congestion, limiting the exploration of placement options and incurring high iteration costs when design changes are needed.
Innovation Solution
A rectilinear-block placement method using a machine learning model to predict the positions of sub-blocks on a chip canvas based on an edge-depth map, optimizing the placement of flexible blocks by generating an edge-depth map and utilizing reinforcement learning to minimize wirelength and improve placement efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If manual placement methods are used by chip designers, then expertise and control are maintained, but the placement process becomes time-consuming and results in suboptimal solutions
Solution Approach 1:
The patent replaces manual mechanical placement operations with an automated machine learning system. The neural network model automatically predicts optimal placement positions for circuit blocks based on input features, eliminating the need for manual designer intervention and significantly reducing placement time while maintaining or improving placement quality.
Solution Approach 2:
The placement system performs self-optimization through the machine learning model that automatically analyzes design requirements and determines optimal block positions without external human input. The system serves itself by using trained models to make placement decisions, reducing dependency on manual expertise while improving efficiency.
2Manufacturing precision
If the number of placement options is increased to improve solution quality, then optimization potential is enhanced, but the complexity and time required for evaluation increases
Solution Approach 1:
The patent transforms the placement optimization problem by changing the approach from exhaustive evaluation of multiple placement options to a direct prediction model. The neural network learns from training data and directly predicts optimal positions, avoiding the combinatorial explosion of evaluating numerous placement configurations and reducing computational complexity.
Solution Approach 2:
The system performs preliminary training of the machine learning model on historical placement data before actual placement tasks. This preliminary action allows the model to learn optimal placement patterns in advance, so during actual placement it can quickly predict positions without needing to evaluate multiple options in real-time, reducing complexity during the placement phase.
Data Source
AI summary
A rectilinear-block placement method includes disposing a first sub-block of each flexible block on a layout area of a chip canvas according to a reference position, generating an edge-depth map relative to first sub-blocks of flexible blocks on the layout area, predicting positions of second sub-blocks of the flexible blocks with depth values on the edge-depth map by a machine learning model, and positioning the second sub-blocks on the layout area according to the predicted positions of the second sub-blocks of the flexible blocks.