Dynamic Arithmetic Circuit Partitioning for Combinatorial Optimization

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current optimization apparatuses for combinatorial optimization problems are inefficient as they require separate configurations for each problem scale and accuracy, leading to suboptimal performance due to fixed hardware resource utilization and increased operation time for problems of different scales or accuracies.

Innovation Solution

The optimization problem arithmetic method dynamically adjusts the partition mode and execution mode of the arithmetic circuit to match the scale and accuracy of the combinatorial optimization problem, allowing for efficient resource allocation and minimizing abnormal results by changing modes during idle states.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If separate configurations are used for each problem scale and accuracy, then the optimization apparatus can handle specific problem sizes, but the device complexity increases and hardware resource utilization becomes inefficient

Engineering Contradiction:

Improvecapability to handle different problem scalesVSAvoidconfiguration complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements dynamic reconfiguration of the arithmetic circuit by dividing it into multiple partitions that can be dynamically activated or deactivated based on the problem scale. The control unit changes the operational state of partitions during idle periods, transforming the static hardware configuration into a dynamic one that adapts to different problem sizes without requiring separate configurations for each scale.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The arithmetic circuit is segmented into multiple partitions, each capable of handling specific problem scales. This segmentation allows the system to activate only the necessary number of partitions based on the current problem requirements, reducing the effective device complexity while maintaining the capability to handle various problem scales through selective partition activation.

Inventive Principle:

Principle #1Segmentation

2Productivity

If fixed hardware resources are utilized, then the arithmetic circuit structure remains simple, but the productivity decreases for problems of different scales

Engineering Contradiction:

Improvearithmetic processing speedVSAvoidhardware resource allocation complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The system dynamically adjusts hardware resource utilization by controlling the activation state of different partitions based on problem scale and accuracy requirements. This dynamic resource allocation maximizes productivity for each specific problem type without requiring a complete redesign of the hardware architecture, as the same physical resources are efficiently utilized through selective activation.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The arithmetic circuit is designed with universal partitions that can handle multiple problem scales when appropriately activated. Each partition serves multiple functions depending on the problem requirements, allowing the system to achieve high productivity across different problem sizes using the same hardware resources rather than dedicated fixed configurations.

Inventive Principle:

Principle #6Universality (Multi-functionality)

3Productivity

If the arithmetic circuit is logically divided into partitions, then the hardware resource utilization improves, but the device complexity increases

Engineering Contradiction:

Improveresource utilization efficiencyVSAvoidcontrol complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The control unit automatically determines the appropriate partition configuration based on the input problem characteristics, eliminating the need for manual or complex external control. The system self-adjusts the activation state of partitions according to the problem scale and accuracy requirements, reducing the effective control complexity while maintaining high resource utilization efficiency.

Inventive Principle:

Principle #25Self-service

4Adaptability or versatility

If modes are changed during idle states, then the adaptability to different problem requirements improves, but the operation time increases due to mode switching

Engineering Contradiction:

Improveadaptation to problem scale and accuracyVSAvoidmode switching time

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The system performs mode switching during idle states before the next problem is processed, ensuring that the arithmetic circuit is already configured optimally when computation begins. This preliminary action eliminates mode switching time from the critical computation path, maintaining high adaptability without adding operational delays.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS11106761B2Optimization problem arithmetic method and optimization problem arithmetic apparatus

Publication Date: 2021.08.31 FUJITSU LTD

AI summary

A computer-implemented optimization problem arithmetic method includes receiving a combinatorial optimization problem, determining, based on scale or a requested accuracy of the combinatorial optimization problem, a partition mode and an execution mode, the partition mode defining a logically divided state of an arithmetic circuit, the execution mode defining a range of hardware resources to be utilized in arithmetic operation for each of partitions generated by logically dividing the arithmetic circuit, and causing the arithmetic circuit to execute arithmetic operation of the combinatorial optimization problem in accordance with the determined partition mode and the determined execution mode.