Configurable GPU Core Allocation for Virtualized Graphics Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing graphics processing unit (GPU) virtualization technologies struggle with fixed and are difficult to flexibly configure graphics processing unit cores used by virtual graphics processing units are fixed and are difficult to flexibly configure graphics processing unit unit cores used by virtual graphics processing unit cores used by virtual graphics processing unit unit units according to according to actual demands, leading to inflexible configuration and potential congestion and increased chip area.
Innovation Solution
A graphics processing unit with configurable graphics processing unit cores and data-and-command dispatchers, allowing flexible configuration of virtual graphics processing units based on demand, optimizing connections to reduce congestion and chip area.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If graphics processing unit cores are fixed in virtual graphics processing units, then the system structure is simple, but the configuration flexibility according to actual demands is poor
Solution Approach 1:
The patent implements dynamic configuration of graphics processing unit cores to virtual graphics processing units through a configurable connection matrix. The connection matrix allows dynamic allocation and reassignment of cores to virtual units based on actual demand, transforming the static fixed assignment into a dynamic flexible configuration system.
Solution Approach 2:
The connection matrix serves as a universal switching mechanism that can accommodate multiple virtual graphics processing units with varying numbers of cores. Each virtual unit can be configured with different core assignments, making the system universally adaptable to different workload requirements without requiring separate dedicated connections for each configuration scenario.
2Adaptability or versatility
If more connection lines are used to connect data-and-command dispatchers to graphics processing unit cores, then the connection flexibility is improved, but the chip area increases and congestion occurs during physical design and routing
Solution Approach 1:
Instead of creating unique dedicated connection lines for each possible virtual graphics processing unit configuration, the patent uses a shared connection matrix that can be logically configured in multiple ways. The physical connection infrastructure is copied or replicated in a modular fashion, allowing the same physical lines to serve multiple virtual configurations through software-defined routing.
Solution Approach 2:
The connection matrix acts as an intermediary layer between the data-and-command dispatchers and the graphics processing unit cores. This intermediate switching structure enables flexible virtual connections without requiring direct dedicated lines for each virtual unit, thereby reducing the total number of physical connection lines needed while maintaining high connection flexibility.
3Adaptability or versatility
If a fixed connection method is used between data-and-command dispatchers and graphics processing unit cores, then the physical design is simple, but the ability to meet different virtual graphics processing unit demands is limited
Solution Approach 1:
The connection system is segmented into modular components: data-and-command dispatchers, a configurable connection matrix, and graphics processing unit cores. This segmentation allows the connection matrix to be designed as a reusable module that can be instantiated in different configurations, simplifying the physical design while enabling versatile virtual unit configurations.
Solution Approach 2:
The patent changes the configuration parameters of the connection matrix to adapt to different virtual graphics processing unit demands. By modifying connection parameters such as routing paths, allocation ratios, and mapping relationships, the system can meet various virtual unit requirements without changing the underlying physical architecture, thus maintaining ease of manufacture.
Data Source
AI summary
The present disclosure provides a graphics processing unit, a chip, and an electronic device. The graphics processing unit includes at least two data-and-command dispatchers and at least two graphics processing unit cores, each data-and-command dispatcher is connected to at least one graphics processing unit core, and one of the data-and-command dispatchers is connected to one of the graphics processing unit core through a set of data-and-command transmission lines; The graphics processing unit is configured to provide at least one virtual graphics processing unit, and each virtual graphics processing unit includes one data-and-command dispatcher and some or all of the graphics processing unit cores connected to the data-and-command dispatcher. The graphics processing unit can provide at least one virtual graphics processing unit, and the graphics processing unit can be shared by multiple users in a manner of virtualization of the graphics processing unit.