Data Accumulation Processor Bit-Range Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing data processing techniques face challenges in efficiently accumulating and storing large volumes of data, particularly in neural networks, due to high data bandwidth requirements when using large storage devices.

Innovation Solution

A processor is designed to divide input data into multiple bit ranges, processing and accumulating data in each range separately. The accumulated data is then bit-shifted and stored in multiple storage devices with different bit ranges, reducing the bandwidth required for data transfer.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If large storage devices are used to accumulate and store large volumes of data, then data accumulation capability is improved, but data bandwidth requirements increase

Engineering Contradiction:
Improvedata accumulation capabilityVSAvoiddata bandwidth requirements
Core Design Contradiction:
Quantity of substanceVSObject-generated harmful factors

Solution Approach 1:

The input data is divided into multiple bit ranges (e.g., first bit range, second bit range, third bit range). Each bit range is processed separately through its own accumulation path, allowing the system to accumulate large volumes of data while transferring only portions (e.g., lowest n bits) to storage devices, thereby reducing the data bandwidth required for transfer while maintaining data accumulation capability

Inventive Principle:
Principle #1Segmentation

2Productivity

If data is processed and accumulated in large volumes, then processing throughput is improved, but storage requirements increase

Engineering Contradiction:
Improveprocessing throughputVSAvoidstorage requirements
Core Design Contradiction:
ProductivityVSVolume of stationary object

Solution Approach 1:

Only the lowest n bits of the accumulated data from each bit range are extracted and transferred to the storage device. The remaining higher bits are retained in the processing system for further accumulation operations. This extraction approach allows the system to maintain high processing throughput while minimizing the storage requirements, as only essential portion of the accumulated data needs to be persisted

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentUS12271608B2Data accumulation
Publication Date: 2025.04.08 ARM LTD
  • US12271608B2 patent drawing
  • US12271608B2 patent drawing
  • US12271608B2 patent drawing

AI summary

A processor to generate accumulated data comprising, for an operation cycle: performing an operation on a first bit range of a set of first input data to generate a set of operation data, which is accumulated with stored data within a first storage device. A lowest n bits of the accumulated data are accumulated with first further stored data within a first bit range of a second storage device, and are bit-shifted from the first storage device. Further accumulated data is generated, comprising, for an operation cycle: performing the operation on a second bit range of the set of first input data to generate a further set of operation data, which is accumulated with the stored data within the first storage device. A lowest m bits of the further accumulated data is accumulated with second further stored data within a second bit range of the second storage device.