Surface Library for Data Sharing in Autonomous Machine Graphics Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques are inefficient for data sharing across processing systems and do not facilitate the expansion or re-expansion of compressed models, limiting performance and communication efficiency in parallel graphics data processing.
Innovation Solution
A novel technique using a surface library allows data produced by one graphics processor to be shared and retrieved by another, and enables the expansion of compressed models for efficient communication, utilizing detection/observation logic, library generation/mapping logic, data sharing/retrieval logic, and compression/expansion logic.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional data sharing techniques are used across processing systems, then data can be transferred between systems, but the process is inefficient and cumbersome
Solution Approach 1:
The patent introduces a surface library as an intermediary data storage structure that enables efficient data sharing between processing systems. The surface library acts as a mediator that stores data in a standardized format, allowing different processing systems to access and share data without direct complex interactions, thereby improving efficiency and reducing operational complexity
Solution Approach 2:
The surface library is designed as a universal data sharing mechanism that can be used across multiple different processing systems and architectures. It provides a common interface and data structure that works universally, eliminating the need for system-specific data sharing implementations and reducing overall complexity
2Quantity of substance
If data is compressed for communication, then bandwidth requirements are reduced, but the system cannot expand or re-expand the compressed models
Solution Approach 1:
The patent implements preliminary compression of data models before transmission or storage in the surface library. By compressing data in advance, the system reduces bandwidth requirements and storage needs while maintaining the ability to expand the data when needed. The compression is performed beforehand, allowing the actual model data to be stored in compact form while preserving full expansion capability through the surface library interface
Solution Approach 2:
The system dynamically changes the compression parameter state of data in the surface library, allowing data to transition between compressed and expanded states as needed. This parameter change capability enables the system to maintain compact storage and transmission while providing on-demand expansion functionality, thus achieving both reduced data volume and maintained adaptability
3Productivity
If parallel processing is implemented to increase throughput, then processing speed improves, but data sharing between parallel processors becomes more complex
Solution Approach 1:
The patent merges the data sharing infrastructure into a unified surface library that serves all parallel processors. Instead of implementing separate data sharing mechanisms for each processor pair, the system combines all data sharing operations through a common library, reducing the overall complexity of the data sharing infrastructure while maintaining high throughput for all parallel operations
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
A mechanism is described for facilitating sharing of data and compression expansion of models at autonomous machines. A method of embodiments, as described herein, includes detecting a first processor processing information relating to a neural network at a first computing device, where the first processor comprises a first graphics processor and the first computing device comprises a first autonomous machine. The method further includes facilitating the first processor to store one or more portions of the information in a library at a database, where the one or more portions are accessible to a second processor of a computing device.