Logic Classification Unit for Programmable Chip Data Routing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The increasing complexity of network services leads to a surge in the number of arithmetic logic units (ALUs) required for processing, resulting in a significant increase in interconnection relationships and connection resources, which in turn raises costs for physical implementation in programmable processing architectures.

Innovation Solution

A data processing method and interaction system that classify data into categories, allowing only corresponding ALUs to access specific data, thereby reducing the number of connection lines and resources needed between ALUs and the data bus, by using a logic classification unit to send targeted data to specific ALUs based on a preset mapping relationship.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If each ALU is interconnected to any output interface of a data bus to enable random access to any data, then processing flexibility and scalability are improved, but the quantity of connection lines and physical implementation costs increase significantly

Engineering Contradiction:
Improveprocessing flexibilityVSAvoidconnection resources
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent segments the data bus into multiple classified data buses, where each classified data bus carries data of a specific category. ALUs are also segmented into different types based on their processing needs. Each ALU is connected only to the classified data buses relevant to its function, rather than being connected to all data buses. This segmentation reduces the connection complexity while maintaining the ability to access required data, resolving the contradiction between flexibility and connection resources.

Inventive Principle:
Principle #1Segmentation

2Productivity

If the number of ALUs is increased to process complex network services, then service processing capability is improved, but the interconnection relationships and connection resources increase exponentially

Engineering Contradiction:
Improveservice processing capabilityVSAvoidinterconnection relationships
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent applies local quality by assigning different connection characteristics to different ALUs based on their specific processing requirements. Each ALU type is connected to specific classified data buses that carry the categories of data it needs to process. This means that not all ALUs have the same connection pattern - instead, each ALU has connections tailored to its local processing needs. This allows the system to scale the number of ALUs for complex services while maintaining manageable connection resources, as each new ALU only requires connections to the specific classified data buses it needs.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP4160396A1Data processing method and interaction system
Publication Date: 2023.04.05 HUAWEI TECH CO LTD
  • EP4160396A1 patent drawingFigure 1~2
  • EP4160396A1 patent drawingFigure 3
  • EP4160396A1 patent drawingFigure 4

AI summary

Embodiments of this application disclose a data processing method and an interaction system, which are applied to a programmable chip. The method specifically includes: A logic classification unit obtains, based on first data received through a data bus, at least first target classified data and second target classified data, and sends the first target classified data and the second target classified data to a corresponding first ALU and a corresponding second ALU based on a preset mapping relationship. The logic classification unit classifies target execution information that is obtained through preprocessing an entry by a TCAM and service data, so that an instruction memory determines first information and second information, and sends the first information and the second information to the corresponding first ALU and the corresponding second ALU. The first ALU and the second ALU respectively send, through the data bus, data obtained through performing calculation based on the first target classified data and the first information and data obtained through performing calculation based on the second target classified data and the second information. In embodiments of this application, only classified data needs to be sent to a corresponding classified ALU, and the corresponding classified ALU is no longer interconnected with all data, so that connection resources may be saved, and costs for physical implementation may be reduced.