Random-Access Image Compression for GPU Memory Bandwidth
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image data compression methods require decompressing entire blocks to access individual values, leading to inefficient data access for rendering units, especially in graphics processing units (GPUs) where memory bandwidth is a limited resource.
Innovation Solution
A lossless compression method that allows individual values in a block to be accessed without decompressing the whole block by using fixed-length compression algorithms to represent each value with common base information and a fixed-length parameter, enabling random access to compressed data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If traditional block-based compression is used, then data compression ratio is improved, but data access efficiency deteriorates because entire blocks must be decompressed to access individual values
Solution Approach 1:
The patent segments the compressed block into multiple sub-blocks, where each sub-block contains compressed data for a specific region. This allows the rendering unit to access only the required sub-block rather than decompressing the entire block, thus improving data access efficiency while maintaining compression ratios.
Solution Approach 2:
The patent introduces an intermediary structure (sub-block level) between the compressed block and individual pixel values. This intermediary allows partial decompression at the sub-block level, serving as a mediator that enables efficient random access without requiring full block decompression.
2Speed
If memory bandwidth is increased to improve rendering performance, then data transfer speed is improved, but power consumption increases
Solution Approach 1:
The patent applies partial action by allowing the rendering unit to decompress only the necessary sub-blocks rather than entire blocks. This reduces the volume of data transferred through the memory subsystem, thereby reducing power consumption while maintaining rendering performance.
3Ease of operation
If fixed-length compression algorithms are used, then random access to compressed data is enabled, but compression ratio may be reduced compared to variable-length algorithms
Solution Approach 1:
The patent segments the image data into multiple sub-blocks, each compressed using fixed-length algorithms. This segmentation allows the system to maintain random access capability at the sub-block level while mitigating the compression ratio penalty by applying compression at a finer granularity.
Data Source
AI summary
Methods and compression units for compressing a block of image data, the block of image data comprising a plurality of image element values, the image element values being divisible into at least a first value and a second value such that the block of image data comprises a two-dimensional block of first values, the method comprising: compressing a first data set comprising all or a portion of the two-dimensional block of first values in accordance with a first fixed-length compression algorithm to generate a first compressed block by: identifying common base information for the first data set; and identifying a fixed-length parameter for each first value in the first data set, the fixed-length parameter being zero, one or more than one bits in length; and forming a compressed block for the block of image data based on the first compressed block.