Dual-Module Decompression for 10-Bit GPU Image Data

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing compression and decompression units for GPUs are inefficient in handling 10-bit data, requiring additional hardware and increasing silicon area, while existing schemes compromise image quality and efficiency.

Innovation Solution

A dual-compression scheme is employed, where 10-bit data is split into most significant bits (MSBs) and least significant bits (LSBs), with separate compression and decompression modules for each, allowing concurrent processing and maintaining high image quality by preserving 9th bit information and minimizing errors in constant regions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If compression and decompression units are configured in hardware to handle 10-bit data values in addition to 8-bit data values, then the capability to compress and decompress 10-bit image data is improved, but the silicon area increases significantly

Engineering Contradiction:
Improvecapability to compress and decompress 10-bit image dataVSAvoidsilicon area
Core Design Contradiction:
Adaptability or versatilityVSArea of stationary object

Solution Approach 1:

The compression and decompression units are designed with multi-functionality to handle both 8-bit and 10-bit data values using the same hardware infrastructure. The units can dynamically adapt their operation mode based on the input data format, eliminating the need for separate dedicated hardware for each data type and thus reducing silicon area while maintaining versatility

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The compression and decompression units utilize parameter changes by adjusting their internal configuration and processing parameters based on whether 8-bit or 10-bit data is being processed. This allows a single hardware unit to efficiently handle different data formats by changing operational parameters rather than requiring separate physical units for each format

Inventive Principle:
Principle #35Parameter changes

2Loss of energy

If data is compressed before transfer to memory, then the memory bandwidth is reduced and power consumption is reduced, but the compression and decompression process requires additional hardware complexity

Engineering Contradiction:
Improvepower consumptionVSAvoidhardware complexity
Core Design Contradiction:
Loss of energyVSDevice complexity

Solution Approach 1:

The compression and decompression functionality is segmented into modular components that can be selectively activated based on the data format and processing requirements. This segmentation allows the system to implement compression with minimal additional hardware complexity by only activating the necessary processing stages rather than requiring a complete separate compression system

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces intermediary processing stages that bridge between the existing 8-bit optimized hardware and the required 10-bit处理能力. These intermediary components enable compression functionality without requiring a complete redesign of the hardware architecture, thus reducing overall hardware complexity while still achieving the power consumption benefits of compression

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP4254805B1Data compression and decompression
Publication Date: 2026.02.18 IMAGINATION TECH LTD
  • EP4254805B1 patent drawingFigure 1A
  • EP4254805B1 patent drawingFigure 1B
  • EP4254805B1 patent drawingFigure 2

AI summary

A computer-implemented method for decompressing compressed data, the compressed data representing an n-bit decompressed data value, the compressed data comprising a first compressed subset of bits and a second compressed subset of bits, the first compressed subset representing the n - 2 most significant bits of the decompressed data value and the second compressed subset representing the two least significant bits of the decompressed data value, the method comprising performing decompression of the first compressed subset using a first decompression module to determine the n - 2 most significant bits of the decompressed data value; performing decompression of the second compressed subset using a second decompression module to determine the two least significant bits of the decompressed data value, the first and second decompression modules implementing different decompression schemes; and combining the determined n - 2 most significant bits of the decompressed data value and the determined two least significant bits of the decompressed data value to determine the n-bit decompressed data value.