Processor-in-Memory Data Pattern Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current memory devices face inefficiencies in processing and power usage when performing text searches or data inquiries, particularly when dealing with multiple documents or pages, due to the separation of processing resources from memory devices, leading to longer delays and increased power consumption.
Innovation Solution
The implementation of a processor-in-memory (PIM) device with a two-dimensional array of memory cells that replicates a data pattern across rows for concurrent comparison, allowing for high-performance searches by executing arithmetic and Boolean logical operations directly within the memory device, reducing the need for external processing and minimizing power usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If processing resources are separated from memory devices, then device functionality is modularized, but processing delays and power consumption increase
Solution Approach 1:
The patent merges processing resources directly into the memory device by integrating a processing structure within the memory array. This allows search operations to be performed concurrently across multiple memory cells without external processing delays, resolving the contradiction between modularization benefits and processing speed requirements.
Solution Approach 2:
The patent introduces a new dimensional approach by replicating search patterns across multiple memory rows and performing parallel comparisons. This multi-row concurrent processing adds a dimension of parallelism that eliminates sequential processing delays while maintaining modular device architecture.
2Device complexity
If processing resources are separated from memory devices, then device design is simplified, but power consumption increases
Solution Approach 1:
By combining memory storage and processing functions into a single integrated structure, the patent eliminates the need for repeated data transfers between separate memory and processing units. This integration maintains design simplicity while dramatically reducing power consumption associated with data movement and external processing operations.
Solution Approach 2:
The memory device performs search operations autonomously using its own integrated processing structure, eliminating the need for external processing resources. This self-service capability reduces power consumption by keeping all operations within the memory device boundaries, avoiding the energy costs of external communication and processing.
3Adaptability or versatility
If data pattern search is performed using external processing resources, then processing flexibility is maintained, but search performance decreases
Solution Approach 1:
The patent achieves high-performance searching by replicating the search pattern across multiple memory rows and performing concurrent comparisons in parallel. This multi-dimensional parallel processing approach maintains processing flexibility while dramatically improving search performance through simultaneous pattern matching across the entire memory array.
Data Source
AI summary
A method includes replicating a data pattern to be searched for, storing the replicated data pattern into a first row of memory cells, and comparing the replicated data pattern to data stored into a second row of memory cells. In response to detecting the data pattern in the data stored into the second row of memory cells, storing into a third row a value indicative of at least one of an occurrence of the data pattern or a position of the searched for data pattern in the data stored into the row of memory cells.