Register Power Gating for Data Processing Systems

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current data processing systems, particularly those using neural networks for image classification, face high power consumption due to the need for continuous processing of single-bit storage devices, even when only a subset of the data is required, leading to inefficiencies in power usage.

Innovation Solution

Implementing a method where single-bit storage devices are placed into a power saving state based on determining the output data size and input data size, using clock gating or other methods to reduce unnecessary processing, thereby limiting the number of active storage devices and conserving power.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Use of energy by moving object

If all single-bit storage devices are kept active for continuous processing, then data processing capability is maintained, but power consumption increases

Engineering Contradiction:
Improvepower consumptionVSAvoiddata processing capability
Core Design Contradiction:
Use of energy by moving objectVSProductivity

Solution Approach 1:

The patent applies dynamics by making the storage device configuration changeable between static and dynamic modes. The system dynamically adjusts the number of active storage devices based on processing needs, transitioning from a fixed all-active configuration to a flexible configuration where only necessary devices remain active, thereby reducing power consumption while maintaining processing capability.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent segments the storage devices into multiple groups or sets, allowing independent control of activation states. By dividing the storage resources into manageable segments, the system can activate only the required portion for each processing task, avoiding the need to keep all devices active simultaneously, thus reducing overall power consumption.

Inventive Principle:
Principle #1Segmentation

2Use of energy by moving object

If the number of active storage devices is reduced to save power, then power consumption decreases, but data storage capacity is limited

Engineering Contradiction:
Improvepower consumptionVSAvoiddata storage capacity
Core Design Contradiction:
Use of energy by moving objectVSQuantity of substance

Solution Approach 1:

The system dynamically adjusts storage capacity allocation based on task requirements. By making the active storage device configuration changeable, the system provides sufficient storage capacity when needed while reducing active devices during low-demand periods, thus resolving the contradiction between power savings and storage availability.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent implements multi-functionality by enabling storage devices to serve different purposes at different times. The same physical storage devices can be activated in different configurations for different processing tasks, allowing the system to provide varying levels of storage capacity as needed without requiring permanently active devices for all possible scenarios.

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

Data Source

PatentUS10545559B2Data processing system and method
Publication Date: 2020.01.28 ARM LTD
  • US10545559B2 patent drawing
  • US10545559B2 patent drawing
  • US10545559B2 patent drawing

AI summary

A data processing system comprising an arithmetic logic unit (ALU) configured to perform a data processing operation. The data processing system also includes a register arranged to receive an output of the data processing operation, the register comprising a plurality of single-bit storage devices arranged to store binary data of a predetermined data size. A control system is arranged to place a subset of the plurality of single-bit storage devices into a power saving state based on at least one parameter relating to the data processing operation, to limit an output data size of the output of the data processing operation to less than the predetermined data size.