Memory Contiguous Arrangement for Film Grain Read Request Reduction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The existing film grain technology requires numerous read requests to access an 8×8-byte block of memory, which is inefficient for high-performance processors like GPUs due to serial storage configuration of film grain images in the film grain database.

Innovation Solution

The memory system arranges portions of memory contiguously, allowing for reduced read requests by treating multiple 4×4-byte portions as a single 8×8-byte block, enabling a single request to retrieve an 8×8-byte block with sequential addresses.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If portions of film grain image are stored serially in memory, then the memory structure is simple and easy to implement, but the number of read requests required to access an 8×8-byte block increases significantly

Engineering Contradiction:
Improvememory structureVSAvoidmemory access efficiency
Core Design Contradiction:
Device complexityVSProductivity

Solution Approach 1:

The patent reorganizes memory storage by merging multiple 4×1-byte portions into contiguous 8×8-byte blocks. Instead of serially storing portions that require multiple separate read requests, the invention combines related memory portions into single contiguous blocks that can be accessed with fewer read requests, thereby improving memory access efficiency while maintaining implementation simplicity

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent changes the dimensional organization of memory storage from a serial one-dimensional arrangement to a two-dimensional block structure. By organizing memory portions into 8×8-byte blocks with specific row and column arrangements, the invention enables more efficient access patterns that reduce the number of read requests required, transforming the memory access problem from sequential to block-based operations

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Reliability

If eight 8-byte read requests are issued to retrieve an 8×8-byte block from serially stored portions, then complete data retrieval is achieved, but the latency increases and system performance decreases

Engineering Contradiction:
Improvedata retrieval completenessVSAvoidmemory access latency
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent applies preliminary action by pre-organizing memory portions into contiguous 8×8-byte blocks before access operations. This pre-arrangement ensures that when data retrieval is needed, the complete 8×8-byte block is already positioned for efficient access, eliminating the need for multiple sequential read requests and thereby reducing access latency while maintaining data retrieval completeness

Inventive Principle:
Principle #10Preliminary action

3Reliability

If multiple read requests are used to access memory portions, then data access completeness is ensured, but the gate count and system complexity increase

Engineering Contradiction:
Improvedata access completenessVSAvoidsystem complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent merges multiple separate read request operations into a single unified access operation by organizing memory into contiguous 8×8-byte blocks. This consolidation reduces the number of individual read requests from eight to one or fewer, thereby reducing system complexity and gate count while ensuring that complete data access is still achieved through the unified block access mechanism

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentUS8233061B1System, method, and computer program product for reducing a number of requests required to retrieve portions of memory
Publication Date: 2012.07.31 NVIDIA CORP
  • US8233061B1 patent drawing
  • US8233061B1 patent drawing
  • US8233061B1 patent drawing

AI summary

A memory system, method, and computer program product are provided. In use, a plurality of portions of memory is arranged contiguously. Further, a number of requests required to retrieve the portions of memory is reduced utilizing the arrangement. In one possible embodiment, such technique may be used in the context of film grain technology (FGT), such that the portions of memory include portions of a film grain image stored in a film grain database (FGDB).