Tree-Structured Accelerator Interfaces for Cross-Vendor Computing Access
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional computing environments face high computing costs and low performance for complex tasks like privacy computing, with hardware accelerator cards from different manufacturers having unique interfaces that complicate access and usage, making it difficult to achieve practical performance.
Innovation Solution
A computing acceleration method using a tree-structured interface set to determine and access target hardware accelerator cards, allowing partial function execution by decomposing computing tasks and utilizing software modules for complete task completion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If hardware accelerator cards from different manufacturers are used for computing acceleration, then computing performance is improved, but interface complexity and access difficulty increase
Solution Approach 1:
The patent introduces a unified access interface that can accommodate hardware accelerator cards from different manufacturers. This universal interface abstracts the underlying hardware differences, allowing a single access method to work with multiple hardware accelerators. The interface set includes root interfaces and child interfaces that can adapt to different hardware configurations, enabling one interface to serve multiple purposes across different device types and manufacturers.
Solution Approach 2:
The patent uses an interface as an intermediary between the computing program and the hardware accelerator card. This interface layer mediates the interaction by providing standardized access methods while hiding the complexity of different hardware implementations. The interface set acts as a mediator that translates high-level computing tasks into hardware-specific operations, eliminating the need for direct interaction with manufacturer-specific interfaces.
2Productivity
If hardware accelerator cards with manufacturer-specific interfaces are used, then computing acceleration capability is improved, but ease of operation and access cost increase
Solution Approach 1:
The unified access interface provides universal functionality that works across different hardware accelerators and manufacturers. Users can access any hardware accelerator card using the same standardized interface, eliminating the need to learn multiple manufacturer-specific operations. The interface set can adapt to different hardware capabilities while maintaining consistent operation methods.
Solution Approach 2:
The patent creates a virtual copy of the hardware accelerator interface through software abstraction. Instead of directly interacting with physical hardware-specific interfaces, the system uses a standardized software interface that replicates the functionality of various hardware accelerators. This virtual interface layer simplifies operation by presenting a consistent, manufacturer-independent API.
3Adaptability or versatility
If computing tasks are decomposed to access hardware accelerator cards, then adaptability to different hardware is improved, but task complexity increases
Solution Approach 1:
The patent segments the interface into a hierarchical structure with root interfaces and child interfaces. Each interface can be independently selected and activated based on the hardware accelerator type. This segmentation allows the system to adapt to different hardware by activating only the necessary interface components, reducing overall complexity while maintaining versatility.
Solution Approach 2:
The interface set is dynamic and can automatically adapt to different hardware configurations. The system can select appropriate interfaces based on the detected hardware accelerator type, and the interface hierarchy can be dynamically adjusted. This dynamic adaptation simplifies task decomposition by automatically matching the right interface to the hardware without requiring manual configuration.
Data Source
AI summary
Embodiments of this specification provide a computing acceleration method and apparatus, and the method includes: determining a target interface corresponding to a target hardware accelerator card from a preset first tree-structured interface set, where the first tree-structured interface set includes a plurality of interfaces, and the plurality of interfaces are respectively corresponding to a plurality of computing functions; the plurality of interfaces include a root interface, a non-root interface other than the root interface in the plurality of interfaces uses another interface in the plurality of interfaces as a parent interface, and a computing function corresponding to each non-root interface is a sub-computing function obtained by decomposing a computing function corresponding to a parent interface of the non-root interface; and accessing the target hardware accelerator card through the target interface, where the target hardware accelerator card is configured to execute a target computing function corresponding to the target interface.