Reconfigurable Circuit Design Generator for Area and Power Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing reconfigurable architectures are overly flexible and inefficient due to fixed homogenous interconnect systems, leading to larger and more power-consuming circuits than necessary, as they are optimized for a family of applications rather than specific tasks, which becomes a drawback as we approach the limits of Moore's Law.
Innovation Solution
A method for generating a circuit design that is over-provisioned based on target data and training data, optimizing the design to meet specific application requirements while minimizing unnecessary flexibility and power consumption, by adding nodes and edges to create a future-proof circuit that can implement additional capabilities without excessive resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a fixed homogenous interconnect system is used to enable wide variety of applications, then adaptability is improved, but circuit area and power consumption increase
Solution Approach 1:
The patent applies local quality by making interconnect resources adaptive rather than fixed and homogenous. Different interconnect resources are dynamically allocated and configured based on specific application requirements, allowing each region to have properties optimized for its intended function rather than using a uniform interconnect structure throughout the circuit.
Solution Approach 2:
The patent implements dynamics by transitioning from a fixed interconnect system to a reconfigurable one. The interconnect resources can be dynamically programmed and reconfigured at runtime to match the specific needs of different applications, enabling the circuit to adapt its connectivity pattern rather than being constrained by a predetermined fixed structure.
2Adaptability or versatility
If a fixed homogenous interconnect system is used to enable wide variety of applications, then adaptability is improved, but power consumption increases
Solution Approach 1:
The patent applies local quality by making interconnect resources adaptive rather than fixed and homogenous. Different interconnect resources are dynamically allocated and configured based on specific application requirements, allowing each region to have properties optimized for its intended function rather than using a uniform interconnect structure throughout the circuit.
Solution Approach 2:
The patent implements dynamics by transitioning from a fixed interconnect system to a reconfigurable one. The interconnect resources can be dynamically programmed and reconfigured at runtime to match the specific needs of different applications, enabling the circuit to adapt its connectivity pattern rather than being constrained by a predetermined fixed structure.
3Manufacturing precision
If manual optimisation is performed for cell types and positions, then manufacturing precision is improved, but device complexity increases
Solution Approach 1:
The patent applies self-service by using automated algorithms and tools to perform the optimisation of cell types, counts, and positions. The system generates and evaluates multiple design candidates automatically, selecting the optimal configuration without requiring extensive manual intervention, thereby reducing design complexity while maintaining or improving manufacturing precision.
Solution Approach 2:
The patent uses copying by generating multiple candidate designs through automated exploration of the design space. These candidate designs are evaluated and compared to identify the optimal configuration, allowing the system to learn from multiple variations and select the best solution without manual trial-and-error for each configuration.
Data Source
AI summary
Systems and methods for designing reconfigurable integrated circuits receive target data and training data; and generate a circuit design for implementing the target data which is over-provisioned with respect to the target data according to the training data.