Pixel Block Compression Modes for Guaranteed GPU Memory Savings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data compression methods for GPUs and other processing units face challenges in reducing memory bandwidth and storage space while maintaining efficient random access and compression ratios, especially when dealing with image data, due to varying block sizes and burst transfer requirements.
Innovation Solution
A lossy data compression method that divides blocks of pixels into sub-blocks, analyzes alpha channel values to select from multiple compression modes, including fixed and variable alpha channel modes, ensuring a guaranteed compression threshold is met, thereby reducing memory bandwidth and storage space.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If lossless compression is used to maintain data quality, then data integrity is preserved, but compression ratio is insufficient and memory bandwidth is not reduced enough
Solution Approach 1:
The patent changes the parameter of compression fidelity by introducing multiple compression modes (lossless and lossy) with adjustable quality levels. The system can switch between these modes based on the specific requirements of different data blocks, allowing optimization of compression ratio while maintaining acceptable data quality for applications where perfect reconstruction is not critical.
Solution Approach 2:
The patent divides the image data into multiple blocks and applies different compression modes to different blocks. Some blocks are compressed using lossless methods to preserve critical data integrity, while other blocks use lossy compression to achieve higher compression ratios. This segmentation allows the system to optimize overall compression performance while maintaining reliability where needed.
2Manufacturing precision
If higher quality rendering algorithms are used, then image quality is improved, but memory bandwidth consumption increases
Solution Approach 1:
The patent changes the parameter of data representation by using compressed formats that require decompression before rendering. This allows the system to store and transfer less data (reducing memory bandwidth) while still being able to reconstruct high-quality images when needed, effectively decoupling storage efficiency from rendering quality.
Solution Approach 2:
The patent performs compression before data is transferred to or stored in memory. By pre-compressing the image data, the system reduces the amount of data that needs to be transferred over the memory bus, thereby reducing memory bandwidth consumption while preserving the ability to reconstruct high-quality images when required for rendering.
3Use of energy by moving object
If data is compressed to reduce memory bandwidth, then power consumption is reduced, but random access capability is degraded
Solution Approach 1:
The patent divides compressed data into independent blocks that can be individually accessed and decompressed. This block-based structure allows the GPU to randomly access specific blocks of compressed data without needing to decompress the entire data set, thereby maintaining random access capability while benefiting from the power savings of compression.
Solution Approach 2:
The patent applies different compression modes to different blocks of data based on local requirements. Blocks that may be randomly accessed can use compression modes that facilitate faster decompression or better access patterns, while blocks that are less likely to be accessed randomly can use more aggressive compression. This local optimization balances power consumption with random access capability.
4Productivity
If block size is increased to improve compression efficiency, then compression ratio improves, but adaptability to varying data patterns decreases
Solution Approach 1:
The patent uses a hierarchical block structure where large blocks are divided into smaller sub-blocks. This allows the system to work with large block sizes for overall compression efficiency while using smaller sub-blocks to adapt to local data patterns and variations. The multi-scale approach enables both high compression ratios and adaptability to different data characteristics.
Solution Approach 2:
The patent introduces dynamic mode selection where the compression mode for each block is determined based on the specific characteristics of that block's data. This dynamic adaptation allows the system to optimize compression ratio for each block individually, accommodating varying data patterns while maintaining overall compression efficiency through the block-based structure.
Data Source
AI summary
Lossy methods and hardware for compressing data and the corresponding decompression methods and hardware are described. The lossy compression method comprises dividing a block of pixels into a number of sub-blocks and then analysing, for each sub-block, and selecting one of a candidate set of lossy compression modes. The analysis may, for example, be based on the alpha values for the pixels in the sub-block. In various examples, the candidate set of lossy compression modes comprises at least one mode that uses a fixed alpha channel value for all pixels in the sub-block and one or more modes that encode a variable alpha channel value.