Hardware Accelerator Universal Driver via Message Structure

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current CPU+FPGA-based hardware acceleration solutions require customized underlying drivers for each service type, leading to poor mobility and flexibility in service acceleration scenarios.

Innovation Solution

A message structure is agreed between the host service layer and the hardware processing unit, allowing transparent transmission of request messages that include acceleration type identifiers, enabling accelerated processing without the need for specialized drivers, and allowing the hardware processing unit to run on different service platforms.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If customized underlying drivers are used for each service type to accelerate FPGAs, then the service layer can achieve accelerated processing for specific service types, but mobility and flexibility are poor and each service type requires a specialized driver

Engineering Contradiction:
Improveaccelerated processing capabilityVSAvoidmobility and flexibility
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent creates a universal underlying driver that can handle multiple service types through a standardized message structure. The message structure includes fields for service type identification and parameters, allowing a single driver implementation to support various acceleration scenarios without requiring customized drivers for each service type.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent uses parameter-based configuration where the message structure contains service type identifiers and parameters that dynamically determine the acceleration behavior. By changing the message parameters rather than the driver code, the system can adapt to different service types, achieving flexibility without sacrificing processing capability.

Inventive Principle:
Principle #35Parameter changes

2Reliability

If specialized underlying drivers are created for each service type, then the service layer can interact with hardware processing units for specific functions, but the system complexity increases and driver coordination becomes necessary

Engineering Contradiction:
Improveservice acceleration functionalityVSAvoiddriver coordination complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent merges the functionality of multiple specialized drivers into a single universal underlying driver. This unified driver handles all service types through a common message structure, eliminating the need for driver coordination and reducing system complexity while maintaining acceleration functionality.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The standardized message structure acts as an intermediary between the service layer and the universal underlying driver. This message structure encapsulates service-specific information in a uniform format, allowing the driver to process different service types without direct service-layer-driver coordination for each specific case.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Adaptability or versatility

If a standardized message structure is used between host service layer and hardware processing unit, then transparency through host driver layer is achieved and mobility is improved, but the message structure design becomes more complex

Engineering Contradiction:
Improvemobility and flexibilityVSAvoidmessage structure design
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The message structure is segmented into distinct fields: message type identifier, service type identifier, parameter fields, and data payload. This segmentation allows for systematic design and processing of different message types while maintaining a consistent overall structure, making the complexity manageable through modular organization.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3291089B1Data processing method and apparatus
Publication Date: 2021.03.17 HUAWEI TECH CO LTD
  • EP3291089B1 patent drawingFigure 1
  • EP3291089B1 patent drawingFigure 2
  • EP3291089B1 patent drawingFigure 3~4

AI summary

Embodiments of the present invention provide a data processing method and apparatus. The data processing method is applied to a scenario in which a hardware processing unit accelerates service data sent by a host. The method is executed by the hardware processing unit, and includes: receiving a request message that is sent from a host service layer and transparently transmitted through a host driver layer, where the request message includes at least one acceleration type identifier and to-be-acceleratedly-processed service data, and each acceleration type identifier corresponds to one type of accelerated processing; and performing at least one type of accelerated processing in a one-to-one correspondence with the at least one acceleration type identifier on the service data. In the method, interaction between the host service layer and the hardware processing unit does not need coordination of a specialized driver, so that dependence on a specific underlying driver for a service layer may be shielded. In this method, the hardware processing unit may run on different service platforms, and have strengthened logic interoperability, thereby improving mobility and flexibility in a service processing process.