Threshold-Based Lossy Color Compression for Graphics Tiles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional graphics processors introduce visual artifacts when compressing color data for tiles with multiple pixel samples, leading to issues like 'cracks' at primitive edges in meshes due to incorrect combination of colors.
Innovation Solution
A system and method using equality and threshold comparisons to compress color data, where equality comparisons are performed on exponent and sign fields or high-order bits, and threshold comparisons are done on mantissa fields or low-order bits, allowing for lossless compression by setting a threshold value to zero.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If color data is compressed by combining color data for multiple samples, then bandwidth requirements are reduced, but visual artifacts are introduced
Solution Approach 1:
The patent applies partial action by performing equality comparison on only a portion of the color data (first portion with significant bits) rather than the entire color data. This allows compression to proceed when the compared portions are equal, while maintaining accuracy by not compressing when differences exist in the remaining portions, thus reducing bandwidth without introducing visual artifacts.
Solution Approach 2:
The patent changes the parameter of comparison by dividing color data into different portions (first portion with exponent/sign fields or high-order bits, and second portion with mantissa fields or low-order bits) and applying different comparison strategies (equality vs. threshold comparison) to each portion. This enables selective compression based on the actual differences present in the data.
2Quantity of substance
If color data is compressed for multiple pixels within a tile, then memory bandwidth is reduced, but cracks appear at primitive edges
Solution Approach 1:
The patent segments the color data into distinct portions (first portion containing exponent and sign fields or high-order bits, and second portion containing mantissa fields or low-order bits). This segmentation allows the system to compare only the significant portions for equality while handling the less significant portions separately, enabling compression without compromising edge precision.
Solution Approach 2:
The patent applies different comparison criteria to different portions of the color data. The first portion (containing significant bits) undergoes strict equality comparison to preserve edge precision, while the second portion (containing less significant bits) can tolerate threshold-based comparison. This local differentiation of quality requirements enables both compression and precision preservation.
3Quantity of substance
If compression is applied to reduce data size, then storage requirements are reduced, but color accuracy is compromised
Solution Approach 1:
The patent performs equality comparison on only the first portion of the color data (containing exponent, sign fields or high-order bits) rather than the complete color data. This partial comparison approach allows the system to identify compressible data while maintaining color accuracy, as the comparison is sufficient to detect meaningful differences without requiring full-data comparison.
Solution Approach 2:
The patent changes the comparison parameter from full-color-data equality to partial-color-data equality combined with threshold comparison. By comparing only the first portion for equality and handling the second portion with threshold comparison, the system achieves compression while preserving color accuracy for significant differences.
Data Source
AI summary
One embodiment of the present invention sets forth a technique for compressing color data. Color data for a tile including multiple samples is compressed based on an equality comparison and a threshold comparison based on a programmable threshold value. The equality comparison is performed on a first portion of the color data that includes at least exponent and sign fields of floating point format values or high order bits of integer format values. The threshold comparison is performed on a second portion of the color data that includes mantissa fields of floating point format values or low order bits of integer format values. The equality comparison and threshold comparison are used to select either computed averages of the pixel components or the original color data as the output color data for the tile. When the threshold is set to zero, only tiles that can be compressed without loss are compressed.