Reorder Unit for Non-Contiguous Pixel Data Buffering

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing systems face challenges in efficiently reordering non-contiguous pixel data from image sensors into contiguous raster order for effective processing, which can lead to increased power consumption and reduced performance.

Innovation Solution

The implementation of a reorder unit with input and output FIFO buffers and memory that reorders non-contiguous pixel data into contiguous order, allowing for efficient data processing without direct CPU involvement, using a pipelined approach and minimal storage resources.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If data reordering is performed using conventional methods, then data can be reordered from non-contiguous to contiguous order, but system performance degrades and energy consumption increases

Engineering Contradiction:
Improvesystem performanceVSAvoidenergy consumption
Core Design Contradiction:
ProductivityVSUse of energy by moving object

Solution Approach 1:

The patent divides the data reordering process into multiple segments by using multiple FIFO buffers (first buffer, second buffer, third buffer, fourth buffer) that handle different portions of the data stream simultaneously. This segmentation allows parallel processing of data packets, improving throughput and reducing the energy cost per unit of work compared to sequential processing methods.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a temporal dimension to the reordering process by using buffered storage to decouple the input data rate from the output data rate. Data is stored in FIFO buffers at different stages, allowing the system to reorder data without requiring simultaneous processing, thus adding time as a dimension to resolve the ordering conflict without performance penalty.

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

2Manufacturing precision

If data reordering operations are performed, then contiguous data ordering is achieved, but general processing resources are occupied and cannot handle other operations

Engineering Contradiction:
Improvedata ordering accuracyVSAvoidprocessing resource availability
Core Design Contradiction:
Manufacturing precisionVSAdaptability or versatility

Solution Approach 1:

The patent introduces FIFO buffers as intermediary storage elements between the data source and the processing resources. These buffers absorb the reordering operations, allowing the main processing resources to receive already-ordered data without having to perform complex reordering algorithms, thus freeing them for other operations while maintaining precise data ordering.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The data reordering system serves itself by using the inherent structure of the data stream and simple FIFO buffer operations rather than requiring external processing resources. The buffers automatically manage the reordering based on data arrival patterns, eliminating the need for CPU intervention and making processing resources available for other tasks.

Inventive Principle:
Principle #25Self-service

3Productivity

If complex reordering operations are implemented, then data can be reordered efficiently, but device complexity increases

Engineering Contradiction:
Improvereordering efficiencyVSAvoidbuffer and memory structure
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent merges multiple simple FIFO buffer operations into a unified reordering system that achieves complex reordering goals. Rather than implementing a single complex reordering algorithm, the system combines several simple buffer stages that collectively perform the reordering function, reducing individual component complexity while maintaining overall efficiency.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP3335107B1Data reordering using buffers and memory
Publication Date: 2020.02.12 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP3335107B1 patent drawingFigure 1
  • EP3335107B1 patent drawingFigure 2
  • EP3335107B1 patent drawingFigure 3

AI summary

Apparatus and methods are disclosed for reordering data received in a non-contiguous order into a contiguous order. In one example of the disclosed technology, an apparatus includes a number of input buffers comprising at least a first, first-in first-out (FIFO) input buffer and a second FIFO input buffer, a number of FIFO output buffers, and a reorder unit configured to store a first portion of non-contiguous data received from an image sensor in the first input buffer, store a second portion of the received data in the second FIFO input buffer, store a respective pixel of data output by the first and second FIFO input buffers at a first address location in the memory, and traverse the memory according to an order to store the respective pixels in a FIFO output buffer. The apparatus can thus be used to reorder pixel data prior to further image processing.