Compressed Surface Flat Mapping for Low-Latency Metadata Access

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing graphics processing systems face increased latency due to the sequential retrieval of compression control data and address translation for compressed surfaces, leading to inefficient access to compressed data in memory.

Innovation Solution

Implement a flat mapping in virtual address space to directly infer the address of compression control data from the address of the data surface, eliminating the need for separate retrieval and reducing latency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If an auxiliary table is used to map between compressed surface and compression control surface, then the compression control data can be stored separately from main data surface, but the access latency increases due to sequential retrieval requirements

Engineering Contradiction:

Improvememory allocation structureVSAvoidaccess latency

Core Design Contradiction:

Device complexityVSLoss of time

Solution Approach 1:

The patent extracts the compression control data from the main data surface and places it in a separate compression control surface (CCS), allowing independent management and optimization of control information while maintaining separate allocation. This extraction enables the system to manage compression metadata independently from the actual compressed data, improving memory utilization while the flat mapping ensures fast access by calculating CCS address directly from data surface address without sequential retrieval

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent introduces a flat mapping mechanism as an intermediary between the data surface and compression control surface. This mapping uses a mathematical relationship (CCS address = data surface address + offset) to directly translate addresses, eliminating the need for auxiliary tables and sequential lookups. The intermediary mapping layer provides direct access paths while maintaining the separation benefits

Inventive Principle:

Principle #24Intermediary (Mediator)

2Adaptability or versatility

If compression control data is stored in separate memory allocation, then memory management becomes more flexible, but address translation complexity increases with potential TLB misses

Engineering Contradiction:

Improvememory management flexibilityVSAvoidaddress translation complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent merges the address translation processes for data surface and compression control surface by establishing a fixed mathematical relationship between their addresses. Instead of requiring separate address translation operations that could each cause TLB misses, the flat mapping allows the system to derive the CCS address directly from the data surface address through simple arithmetic, combining the translation logic and eliminating sequential TLB lookup operations

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent changes the addressing parameter relationship between data surface and compression control surface from a complex mapping (requiring auxiliary tables) to a simple linear relationship (address + offset). This parameter change transforms the address translation from a multi-step process with potential TLB misses into a direct calculation, reducing translation complexity while maintaining separate memory allocation flexibility

Inventive Principle:

Principle #35Parameter changes

3Reliability

If sequential access to compression control data is required before accessing compressed surface, then data integrity is ensured, but access speed decreases

Engineering Contradiction:

Improvedata access integrityVSAvoidaccess speed

Core Design Contradiction:

ReliabilityVSSpeed

Solution Approach 1:

The patent implements preliminary action by pre-establishing the flat mapping relationship between data surface and compression control surface addresses. This mapping is configured in advance with a fixed offset, so when access is needed, the system can immediately calculate the CCS address without sequential retrieval. The preliminary setup of the mapping relationship ensures data integrity while enabling direct access

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent allows the system to skip the sequential retrieval step by using direct address calculation through flat mapping. Instead of following the traditional sequence of looking up auxiliary tables and retrieving control data before accessing the compressed surface, the system rushes through to the target address by calculating it directly from the data surface address, maintaining integrity through the established mapping relationship while significantly improving access speed

Inventive Principle:

Principle #21Skipping (Rushing through)

Data Source

PatentUS12518337B2Compression using a flat mapping in virtual address space

Publication Date: 2026.01.06 INTEL CORP

AI summary

Described herein is a graphics processor comprising a processing resource configured to perform processing operations, a codec configured to compress and decompress data associated with the processing operations, and circuitry configured to calculate a metadata address for a compressed surface based on a flat virtual memory address mapping between the address of the compressed surface and the metadata address. The compressed surface is to store data associated with a processing operation to be performed by the processing resource and the metadata address is a virtual address that stores compression metadata for the compressed surface. The circuitry can configure the codec to access the compressed surface based on the compression metadata.