Accuracy Configurable Approximate Adder with Dynamic Error Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing approximate adder designs are limited in their ability to dynamically adjust accuracy according to changing requirements, as they are typically optimized for a maximum accuracy threshold, which reduces their benefits in contexts where accuracy needs vary over time.
Innovation Solution
An accuracy-configurable approximate adder that can adaptively operate in both accurate and inaccurate modes, with error detection and correction capabilities, and a pipelined architecture that allows for selectable stages and power-gated error correction to conserve power and improve throughput.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If approximate adder designs are optimized for maximum accuracy threshold, then accuracy is improved, but throughput and power consumption benefits are reduced
Solution Approach 1:
The adder design dynamically switches between approximate and accurate modes based on runtime accuracy requirements. The system includes configurable accuracy modes where the adder can operate in approximate mode for high throughput or switch to accurate mode when precision is required, making the performance characteristics adaptable rather than fixed
Solution Approach 2:
The invention changes the accuracy parameter of the adder runtime by configuring different accuracy modes. The adder can be set to produce results with different accuracy levels (e.g., 1-bit accuracy, 2-bit accuracy, or full accuracy) depending on the application requirements, allowing the system to optimize the accuracy-throughput tradeoff dynamically
2Measurement precision
If approximate adder designs are optimized for maximum accuracy threshold, then accuracy is improved, but power consumption increases
Solution Approach 1:
The adder design dynamically switches between approximate and accurate modes based on runtime accuracy requirements. The system includes configurable accuracy modes where the adder can operate in approximate mode for low power consumption or switch to accurate mode when precision is required, making the power consumption characteristics adaptable rather than fixed
Solution Approach 2:
The invention changes the accuracy parameter of the adder runtime by configuring different accuracy modes. The adder can be set to produce results with different accuracy levels (e.g., 1-bit accuracy, 2-bit accuracy, or full accuracy) depending on the application requirements, allowing the system to optimize the power consumption-accuracy tradeoff dynamically
3Measurement precision
If error detection and correction is enabled, then accuracy is improved, but device complexity increases
Solution Approach 1:
The adder is divided into multiple independent sub-adders that can be configured to different accuracy levels. Each sub-adder processes a specific portion of the input and can be enabled or disabled based on the required accuracy, allowing error correction to be applied selectively to only the necessary portions of the computation
Data Source
AI summary
A preferred method of accuracy configuration with an approximate adder receives two input operands and generates a first approximate adder output with a plurality of sub-adders having a first accuracy under a first condition. Error detection and correction is selectively enabled to generate a next approximate adder output having a second accuracy that is higher than the first accuracy under a second condition. In preferred embodiments, a pipelined architecture provides selectable stages and the enablement of each successive stage provides a high level of accuracy. Power gated control can achieve enablement of error correction stages to conserve power.