Multicore GPU State Caching for Selective Rendering Task Distribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing graphics processing units (GPUs) face inefficiencies in processing image rendering tasks due to redundant transmission of state information across cores, leading to increased competition for communication resources and reduced performance.
Innovation Solution
A multicore GPU architecture with a master unit that distributes image rendering tasks and state information selectively to individual cores, ensuring each core receives only the necessary state information for its assigned subset of tasks, thereby reducing unnecessary data transmission and competition for communication resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If state information is transmitted to all cores in a multicore GPU, then all cores have access to complete state data, but redundant transmission increases communication competition and reduces performance
Solution Approach 1:
The patent segments state information into different categories (cumulative state information and subset state information) and selectively transmits only the necessary portions to each core based on its assigned task subset. This resolves the contradiction by maintaining reliable state availability where needed while reducing redundant transmission to improve overall productivity.
Solution Approach 2:
Each core receives customized state information tailored to its specific task assignment rather than a uniform complete state set. The master unit determines which cores need cumulative state information and which need only subset state information, optimizing communication efficiency while ensuring each core has the necessary state data for its assigned tasks.
2Ease of operation
If a central hub distributes tasks to all cores, then task distribution is simplified, but communication resources become congested and idle time increases
Solution Approach 1:
The master unit performs preliminary classification of tasks into subsets and pre-determines which cores should receive which subsets before transmission. This preliminary organization allows efficient targeted communication, reducing communication competition and minimizing core idle time while maintaining operational simplicity through automated task distribution.
3Loss of information
If complete state information is transmitted to each core, then data completeness is ensured, but communication bandwidth is wasted on redundant data
Solution Approach 1:
The patent extracts only the necessary state information portions (cumulative state information for some cores, subset state information for others) rather than transmitting complete state data to all cores. This extraction approach ensures information completeness for each core's needs while significantly reducing communication energy consumption and bandwidth utilization.
Data Source
AI summary
A set of image rendering tasks and state information are distributed in a graphics processing unit (GPU) having a plurality of cores. A first master unit in one of the cores receives the set of image rendering tasks and the state information, and stores the state information in a memory. The first master unit splits the set of image rendering tasks into a first subset of tasks and a second subset of tasks, wherein the first subset of tasks is assigned to the first core, and the second subset of tasks is assigned to the second core. At least a first portion of the state information is transmitted to the first core, and at least a second portion of the state information is transmitted to the second core. The first subset of tasks is transmitted to the first core, and the second subset of tasks is transmitted to the second core.