GPU Frame Buffer Scheduling for Multi-App Image Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cloud server systems face inefficiencies in generating and delivering multiple images corresponding to multiple applications, leading to increased costs and delays due to the limitations of GPU processing and context switching overheads.
Innovation Solution
A data processing system that employs time-divided GPU processing and synchronized image generation and compression coding to optimize the generation and delivery of images for multiple applications, utilizing a single GPU and compression coder to handle multiple applications efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a server generates images for multiple applications in parallel using a single GPU, then resource utilization improves, but processing delays increase due to context switching overhead
Solution Approach 1:
The patent segments the processing of multiple applications by dividing the frame buffer into separate buffer regions for each application. This allows the GPU to process images for different applications sequentially without context switching overhead, as each application's image data is stored in a dedicated region. The segmentation enables efficient resource utilization while avoiding the time loss associated with context switching between applications.
2Quantity of substance
If multiple images are stored in a single frame buffer, then memory usage improves, but reading and coding efficiency deteriorates due to simultaneous access requirements
Solution Approach 1:
The patent applies local quality by assigning different buffer regions within the frame buffer to different applications based on their specific requirements. Each application's image data is stored in a dedicated local region, allowing the coding section to read and process each application's images independently and efficiently. This regional differentiation maintains high memory utilization while enabling optimized access patterns for each application, thereby preserving reading and coding efficiency.
Data Source
AI summary
Methods and apparatus for: generating a plurality of images corresponding to a plurality of applications executed in parallel and storing the plurality of images in a frame buffer; reading out the plurality of images corresponding to the plurality of applications from the frame buffer at different timings for each application and coding the plurality of images that have been read out; and when the reading and coding completes coding of an image corresponding to a first application and initiates coding of an image corresponding to a second application, saving context regarding the first application from the reading and coding to a storage section and setting context regarding the second application from the storage section in the reading and coding, without an intermediary of processing of software that manages the plurality of applications.