Multiple Level Spine Routing for ICs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing one-level spine routing techniques for integrated circuits are sub-optimal, as they are performed one net at a time, leading to inefficient routing and sub-optimal results in terms of total wire lengths, propagation delays, and skew times.
Innovation Solution
Implementing multiple-level spine routing, where pins are partitioned into groups and assigned to both main and second-level spines using statistical algorithms like k-means clustering, to optimize spine assignments and reduce total wire lengths, propagation delays, and skew times.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If one-level spine routing is performed one net at a time, then the routing process is simple and easy to implement, but the routing quality deteriorates with sub-optimal wire lengths, propagation delays, and skew times
Solution Approach 1:
The patent divides the routing process into multiple levels: first-level spines are assigned to individual nets, while second-level spines group multiple nets together. This segmentation allows the system to maintain the simplicity of individual net routing while achieving optimization through hierarchical grouping, resolving the contradiction between ease of implementation and routing quality.
Solution Approach 2:
The patent introduces a hierarchical dimension to traditional spine routing by adding second-level spines that operate at a higher abstraction level. This dimensional change enables simultaneous optimization of multiple nets through statistical algorithms like k-means clustering, transforming the single-level sequential process into a multi-level system that achieves both simplicity and quality.
2Manufacturing precision
If multiple-level spine routing is implemented with statistical algorithms, then routing quality improves with optimized wire lengths and delays, but the computational complexity increases
Solution Approach 1:
The patent segments the computational task into distinct levels: first-level spine assignment for individual nets and second-level spine grouping for multiple nets. This segmentation allows statistical algorithms to be applied selectively at appropriate levels, improving routing quality while managing computational complexity through hierarchical decomposition of the optimization problem.
Solution Approach 2:
The patent performs preliminary actions by first assigning first-level spines to nets before grouping them into second-level spines. This preliminary assignment simplifies the subsequent statistical clustering operation, as the k-means algorithm operates on pre-organized data structures, reducing overall computational complexity while maintaining routing quality improvements.
3Length of moving object
If nets are grouped into second-level spines, then total wire lengths are reduced, but the routing process becomes more complex
Solution Approach 1:
The patent merges multiple first-level spines into second-level spines that serve multiple nets simultaneously. This merging reduces total wire length by sharing common routing resources across nets with similar trajectories, while the hierarchical structure maintains process manageability by organizing complexity at different levels rather than requiring simultaneous optimization of all nets.
Data Source
AI summary
A computer implemented method for routing a net includes selecting, using one or more computer systems, a first spine routing track from a first multitude of routing tracks in accordance with a first cost function, and further in accordance with data associated with the net and the first multitude of routing tracks. The method further includes generating, using one or more computer systems, a first spine wire on the selected first spine routing track.