Incremental NoC Synthesis Tool for Deadlock-Free Network Topologies

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Designing an optimal network-on-chip (NoC) that avoids routing and message-dependent deadlocks while minimizing power and area overhead is challenging, especially when introducing new connections, as current methods require complex and time-consuming redesigns in response to changes in performance requirements or chip floorplan modifications.

Innovation Solution

A tool that generates NoC topology incrementally, synthesizing new connections one at a time, reusing existing segments, and supporting regular network topologies like Mesh networks, with mathematical optimization techniques to determine the synthesis order and ensure deadlock-free routes, while preserving existing routes and minimizing latency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If traditional NoC design methods are used to accommodate new connections, then connectivity requirements are met, but the design process becomes time-consuming and requires frequent redesigns

Engineering Contradiction:
ImproveconnectivityVSAvoiddesign time
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent implements a dynamic incremental synthesis approach where the NoC topology is not fixed but can be continuously modified by adding new connections. The system dynamically updates the topology by synthesizing new connections one at a time while preserving existing routes, allowing the design to adapt to changing requirements without complete redesign.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent performs preliminary actions by establishing a foundation of existing connections and routes before adding new ones. The incremental synthesis methodology prepares the topology in advance by maintaining a set of existing connections that serve as the basis for future expansions, reducing the time needed for subsequent modifications.

Inventive Principle:
Principle #10Preliminary action

2Reliability

If complex routing is implemented to avoid deadlocks, then deadlock-free operation is achieved, but power and area overhead increase

Engineering Contradiction:
Improvedeadlock-free operationVSAvoidpower overhead
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent segments the routing problem by treating each connection synthesis independently while maintaining global deadlock-free properties. By dividing the topology into manageable connection units and synthesizing them incrementally, the system achieves deadlock avoidance without requiring complex global routing schemes that would increase power and area overhead.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary synthesis methodology that mediates between connectivity requirements and deadlock avoidance. The incremental synthesis process acts as an intermediary mechanism that automatically ensures deadlock-free operation while minimizing the complexity of routing paths, thereby reducing power and area overhead compared to traditional complex routing schemes.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Adaptability or versatility

If comprehensive NoC redesign is performed to meet new requirements, then performance requirements are fulfilled, but production delays occur

Engineering Contradiction:
Improveperformance requirementsVSAvoidproduction speed
Core Design Contradiction:
Adaptability or versatilityVSProductivity

Solution Approach 1:

The system transitions from static comprehensive redesign to dynamic incremental updates. Each new connection is synthesized independently and integrated into the existing topology, allowing the design to evolve continuously without stopping production or performing time-consuming complete redesigns.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent recovers and preserves existing valid connections and routes when adding new ones. Instead of discarding the entire design for redesign, the incremental synthesis methodology recovers the existing topology structure and builds upon it, significantly reducing the time and resources needed to meet new performance requirements.

Inventive Principle:
Principle #34Discarding and recovering

4Adaptability or versatility

If new connections are added to existing NoC, then connectivity is enhanced, but routing deadlocks may be introduced

Engineering Contradiction:
ImproveconnectivityVSAvoidrouting deadlock freedom
Core Design Contradiction:
Adaptability or versatilityVSReliability

Solution Approach 1:

The patent applies preliminary anti-action by proactively preventing deadlock formation during the connection synthesis process. The incremental synthesis methodology incorporates deadlock avoidance checks and constraints before new connections are fully integrated, counteracting potential deadlock issues before they can manifest in the enhanced connectivity structure.

Inventive Principle:
Principle #9Preliminary anti-action

Data Source

PatentEP4492278A1Tool for supporting use of regular network topologies in generating a network-on-chip topology
Publication Date: 2025.01.15 ARTERIS INC
  • EP4492278A1 patent drawingFigure 1A
  • EP4492278A1 patent drawingFigure 1B
  • EP4492278A1 patent drawingFigure 2A

AI summary

A tool is disclosed for using custom subnetwork description during generation and synthesis of the network, such as a network-on-chip (NoC). The tool allows for incremental synthesis and transformation of a deadlock-free NoC. The NoC topology is translated into an existing segment; reusing the existing segment in a new route and generating the deadlock-free NoC topology. The tool includes a machine learning model that is trained for synthesis and generation of the NoC and is capable of providing incremental synthesis. The model can also receive feedback from past or previous synthesis for further training of the model.