Processor-in-Memory Query Acceleration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing memory systems face inefficiencies in data query processing due to the need for external circuitry to transfer and process data, leading to delays and bandwidth bottlenecks, which can be exacerbated by the need to process all data rather than only the relevant data matching a query.
Innovation Solution
Implementing processor-in-memory (PIM) capabilities within the memory device to execute data queries internally, allowing for arithmetic operations to be performed on the memory array without transferring data externally, thereby identifying and sending only the relevant data that matches the query.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data is transferred externally for processing, then processing can be performed by external circuitry, but data transfer time and bandwidth usage increase
Solution Approach 1:
The patent combines the processing circuitry with the memory array by implementing processor-in-memory (PIM) capabilities directly within the memory device. This merging eliminates the need for external data transfer by performing query operations directly on the stored data, thereby resolving the contradiction between processing capability and data transfer time.
Solution Approach 2:
The patent introduces an intermediary processing circuitry component that bridges the memory array and external systems. This intermediary performs data processing operations within the memory device itself, acting as a mediator that prevents full data transfer to external circuitry while still enabling complex query operations.
2Measurement precision
If all data is processed to answer a query, then complete accuracy is achieved, but processing time and resource usage increase
Solution Approach 1:
The patent extracts and processes only the relevant data portions that match query criteria directly within the memory array, rather than processing all data. The processing circuitry identifies and extracts matching records, eliminating the need to process unrelated data and thereby improving both accuracy and efficiency.
Solution Approach 2:
The patent applies partial action by performing processing operations only on the specific data portions needed to answer the query. Instead of exhaustive processing of all data, the system performs targeted operations on relevant data segments, achieving sufficient accuracy with reduced processing overhead.
Data Source
AI summary
The present disclosure includes apparatuses and methods for acceleration of data queries in memory. A number of embodiments include an array of memory cells, and processing circuitry configured to receive, from a host, a query for particular data stored in the array, execute the query, and send only the particular data to the host upon executing the query.