Power Gating for Non-Volatile Pooling Storage in CNN Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The computation processing device for convolutional neural networks faces high power consumption, which is a concern for applications in robots, vehicles, and mobile terminals, where energy efficiency is crucial.
Innovation Solution
The device incorporates a convolutional computation unit, a pooling processing unit with a non-volatile storage circuit, and a power gating unit that blocks power supply to the storage circuit when waiting for input data, allowing for reduced power consumption by minimizing current leakage during idle periods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If power supply is continuously provided to the non-volatile storage circuit for pooling, then data retention reliability is improved, but power consumption increases
Solution Approach 1:
The power gating unit periodically controls the power supply to the non-volatile storage circuit for pooling. Power is supplied only during active periods when convolutional computation result data needs to be stored or retrieved, and power is cut off during idle periods when the pooling processing unit is waiting for input data. This periodic on-off power supply pattern reduces overall power consumption while maintaining data retention reliability during active operation through the non-volatile nature of the storage circuit.
2Speed
If the pooling processing unit waits for input data without power gating, then response readiness is improved, but current leakage increases
Solution Approach 1:
The power gating unit implements periodic power supply control to the non-volatile storage circuit for pooling. During idle periods when the pooling processing unit is waiting for convolutional computation result data, power supply is blocked to prevent current leakage. When data arrives, power is restored to enable rapid data storage and processing. The non-volatile storage circuit maintains data integrity during power-off periods, enabling this periodic operation without data loss.
Data Source
AI summary
A computation processing device includes: a convolutional computation unit that sequentially outputs convolutional computation result data; a pooling processing unit including a pooling computation circuit and a non-volatile storage circuit for pooling, in which the non-volatile storage circuit for pooling retains the convolutional computation result data or a computation result of the pooling computation circuit, as retained data, and the pooling computation circuit calculates and outputs pooling data subjected to pooling processing to a pooling region by using the retained data each time when the convolutional computation result data is input from the convolutional computation unit; and a power gating unit that blocks power supply to the non-volatile storage circuit for pooling while waiting for the input of the convolutional computation result data from the convolutional computation unit.