Analog CAM Array Compression by Merging Similar Rows
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Content addressable memories (CAMs) are large, power-consuming, and expensive, limiting their applicability to certain applications despite their efficiency and speed.
Innovation Solution
A method for compressing an analog CAM array by identifying and merging or removing similar rows based on average measures of similarity calculated from discharge currents, reducing the array size by up to 15% without significant loss in model inference accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the analog CAM array size is increased to store more data, then the model inference accuracy is improved, but the power consumption and area increase
Solution Approach 1:
The patent merges similar rows in the analog CAM array by identifying rows with comparable discharge current characteristics and consolidating them into fewer representative rows. This reduces the total number of rows from N to M (where M < N), directly decreasing the array area and power consumption while preserving the essential functionality for model inference through careful selection of representative rows that maintain accuracy.
2Measurement precision
If the analog CAM array size is increased to store more data, then the model inference accuracy is improved, but the area occupied increases
Solution Approach 1:
The patent merges similar rows in the analog CAM array by identifying rows with comparable discharge current characteristics and consolidating them into fewer representative rows. This reduces the total number of rows from N to M (where M < N), directly decreasing the array area and power consumption while preserving the essential functionality for model inference through careful selection of representative rows that maintain accuracy.
3Measurement precision
If random input data is applied to each row to calculate similarity measures, then the compression accuracy is improved, but the computation time increases
Solution Approach 1:
The patent applies partial action by using a limited set of random input data vectors rather than exhaustively testing all possible inputs. This selective sampling approach calculates similarity measures for a representative subset of input cases, achieving sufficient compression accuracy without the prohibitive computational cost of complete enumeration, thus balancing accuracy and computation time effectively.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Reduces the size and power consumption of the CAM array while maintaining accuracy, allowing larger models to be represented and improving operational efficiency.
Implementation Method 1
A mismatch is indicated when a discharge current flows through a pull-down transistor of the analog CAM cell
Data Source
AI summary
A technique for compressing an analog content addressable memory (CAM) array is provided. Random input data is applied to the analog CAM array, and an average measure of similarity is calculated for each output row of the analog CAM array. Rows of the analog CAM array that have measures of similarity that are close to each other can be eliminated, such as by removing similar rows or merging together similar rows. Thus, the analog CAM array size can be reduced without a loss in accuracy of a model stored on the analog CAM array.