Macro Cell Placement Refinement via Simulated Evolution

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional macro placement methods, especially for systems-on-a-chip (SoC), are inefficient and time-consuming due to the complexity introduced by larger macros, leading to longer wirelengths and routing congestion, and rely heavily on manual placement by experienced engineers, which is impractical with increasing macro counts.

Innovation Solution

A simulated-evolution-based macro refinement method that evaluates macro cell scores, uses a predetermined condition with a random number generator to determine placement in a queue, and sorts macros based on scores, incorporating a score function that adjusts wirelength and distance influences over iterations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If simulated annealing algorithm is used for macro placement, then placement can be automated, but execution time becomes excessively long

Engineering Contradiction:

Improvemacro placement automationVSAvoidexecution time

Core Design Contradiction:

Extent of automationVSLoss of time

Solution Approach 1:

The patent divides macro placement into two distinct phases: legalization phase (using simulated annealing for automated placement) and refinement phase (using corner-stitching for precise positioning). This segmentation allows the time-consuming automated placement to be separated from the precision-critical refinement, reducing overall execution time while maintaining automation benefits.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary legalization placement using simulated annealing to establish initial macro positions before conducting corner-stitching refinement. This preliminary action provides a starting configuration that reduces the search space for subsequent refinement, thereby decreasing total execution time while maintaining automation.

Inventive Principle:

Principle #10Preliminary action

2Extent of automation

If conventional simulated annealing method is used, then automation is achieved, but wirelength optimization is insufficient

Engineering Contradiction:

Improveplacement automationVSAvoidwirelength optimization

Core Design Contradiction:

Extent of automationVSManufacturing precision

Solution Approach 1:

The patent segments the placement process into legalization phase (automation-focused) and refinement phase (precision-focused). The corner-stitching refinement specifically targets wirelength optimization by adjusting macro positions along standard cell boundaries, thereby improving manufacturing precision without sacrificing automation.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces corner-stitching as an intermediary refinement step between initial automated placement and final placement. This intermediary process specifically optimizes wirelength by positioning macros to align with standard cell boundaries, thereby improving precision while maintaining the automated workflow.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Manufacturing precision

If manual placement by experienced engineers is used, then placement quality is high, but productivity decreases with increasing macro counts

Engineering Contradiction:

Improveplacement qualityVSAvoidplacement efficiency

Core Design Contradiction:

Manufacturing precisionVSProductivity

Solution Approach 1:

The patent implements self-service through automated legalization and corner-stitching refinement algorithms that automatically optimize macro placement without requiring experienced engineers. The system serves itself by performing both automated placement and precision refinement, thereby maintaining high placement quality while dramatically improving productivity for large-scale macro counts.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent changes the approach from manual parameter tuning by engineers to automated parameter optimization through corner-stitching algorithms. The algorithm automatically adjusts macro positions based on wirelength and routing congestion metrics, achieving expert-level placement quality at automated speed, thus resolving the productivity-precision contradiction.

Inventive Principle:

Principle #35Parameter changes

4Speed

If macros are placed without refinement, then execution speed is fast, but routing congestion increases

Engineering Contradiction:

Improveplacement speedVSAvoidrouting congestion

Core Design Contradiction:

SpeedVSObject-generated harmful factors

Solution Approach 1:

The patent segments placement into a fast legalization phase and a refinement phase. The corner-stitching refinement specifically addresses routing congestion by repositioning macros to align with standard cell boundaries and optimize spacing, thereby reducing harmful routing congestion effects while maintaining overall execution speed through efficient algorithm design.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS10769341B1Method of placing macro cells and a simulated-evolution-based macro refinement method

Publication Date: 2020.09.08 HIMAX TECH LTD

AI summary

A simulated-evolution-based macro refinement method includes evaluating a score of each placed macro cell to be refined; generating a random number; determining whether the score satisfies a predetermined condition; placing the macro cell into a queue if the score associated with the macro cell satisfies the predetermined condition; and sorting and placing macro cells of the queue according to scores of the macro cells in the queue.