In-Memory Computing Circuit Resolving Data Transfer Bottleneck
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The separation between computing/processing units and memory in traditional computer architectures leads to inefficiencies, particularly in data processing speed and energy consumption, as the memory unit's read/write speed lags behind the computing unit's performance, necessitating a solution to integrate computing within the memory unit.
Innovation Solution
An in-memory computing circuit comprising an initial computing circuit and a target computing circuit that perform operations on data to produce results equivalent to those of a half adder, enabling XOR and AND operations directly within the memory, thereby enhancing data processing speed and reducing energy consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data is frequently moved between processor and memory, then computing performance is maintained, but data transfer time and energy consumption increase
Solution Approach 1:
The patent merges the computing circuit and storage circuit into a single integrated memory device. The computing circuit performs computational operations directly on data stored in the storage circuit, eliminating the need for frequent data transfers between separate processor and memory units. This integration resolves the contradiction by combining previously separate functions into one unified system.
Solution Approach 2:
The integrated memory device performs multiple functions: it stores data in the storage circuit and simultaneously performs computing operations in the computing circuit. This multi-functionality allows the same device to both hold data and process it, reducing the need for separate data transfer operations while maintaining high computing performance.
2Productivity
If data is frequently moved between processor and memory, then computing operations can be performed, but energy consumption increases
Solution Approach 1:
By merging the computing circuit and storage circuit into one integrated device, the patent eliminates energy-wasting data transfers between separate components. The computing circuit accesses data directly from the storage circuit within the same device, significantly reducing the energy consumption associated with data movement while maintaining computing performance.
3Device complexity
If separate computing and memory units are used, then functional clarity is maintained, but read/write speed becomes a bottleneck
Solution Approach 1:
The patent combines separate computing and memory functions into a single integrated memory device with both storage and computing circuits. This merger eliminates the speed bottleneck caused by data transfers between separate units, as data can be processed in-place without leaving the storage circuit.
Data Source
AI summary
An in-memory computing circuit includes an initial computing circuit and a target computing circuit. Herein, the initial computing circuit is configured to perform first operation processing on first data and second data to output a first operation result, and perform second operation processing on the first data and the second data to output a second operation result. The target computing circuit is configured to perform the first operation processing on the second operation result and the first operation result to output a first target result, and perform the second operation processing on the first data and the second operation result to output a second target result.