Cross-Subtile Prediction for Low-Latency Bandwidth Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current bandwidth compression techniques for graphics and data processing, particularly in subtile-based processing, face a tradeoff between high throughput and low latency, often resulting in reduced coding efficiency due to the lack of utilization of neighboring pixels with higher correlation during prediction, leading to increased latency and inefficiency.
Innovation Solution
The proposed solution involves utilizing neighboring pixels with higher correlation during the prediction process to enhance coding efficiency, allowing for dependent subtile and pixel predictions, reducing latency, and maintaining high throughput and low latency by leveraging input/output reordering for decoding processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If subtile-based parallel encoding and decoding is used to achieve high throughput, then processing speed is improved, but coding efficiency is reduced due to lack of utilization of neighboring pixels with higher correlation
Solution Approach 1:
The image data is divided into multiple subtiles that can be processed in parallel. Each subtile is further divided into prediction units that can be independently encoded and decoded, enabling high throughput while maintaining the ability to use neighboring pixel correlations within each prediction unit.
Solution Approach 2:
The patent performs preliminary arrangement of data for the set of data subunits into a specific data order before encoding. This preliminary ordering enables efficient parallel processing while preserving spatial correlations, allowing the system to achieve both high throughput and low latency by preparing data in an optimal sequence for subsequent parallel operations.
2Loss of time
If self-contained pixel prediction is used to reduce latency, then processing time is reduced, but coding efficiency is reduced due to no prediction dependency between subtiles
Solution Approach 1:
The patent applies different prediction strategies to different regions within subtiles. By using local quality assessment, the system can utilize neighboring pixel correlations where available while maintaining self-contained prediction where it provides better latency performance, thus optimizing the balance between coding efficiency and latency reduction.
Solution Approach 2:
The data arrangement step performs preliminary organization of pixel data to facilitate both self-contained prediction and cross-subtile correlation utilization. This preliminary action enables the system to achieve low latency through self-contained prediction while maintaining coding efficiency by preserving spatial relationships in the arranged data structure.
Data Source
AI summary
Aspects presented herein relate to methods and devices for data processing including an apparatus, e.g., a CPU. The apparatus may obtain an indication of a set of data subunits corresponding to at least one data unit. The apparatus may also arrange data for the set of data subunits into a first data order for the set of data subunits. Further, the apparatus may perform at least one of an encoding process or a decoding process on the data for each data subunit of the set of data subunits. The apparatus may also rearrange the data for the set of data subunits into the first data order for a first data subunit in the set of data subunits and into a second data order for at least one second data subunit in the set of data subunits, where the first data order is different from the second data order.